Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Hierarchical Phoneme Classification for Improved Speech Recognition

Hierarchical Phoneme Classification for Improved Speech Recognition applied sciences Article Hierarchical Phoneme Classification for Improved Speech Recognition 1,2 3 4 2,5, Donghoon Oh , Jeong-Sik Park , Ji-Hwan Kim and Gil-Jin Jang * SK Holdings C&C, Gyeonggi-do 13558, Korea; odh1501@gmail.com School of Electronic and Electrical Engineering, Kyungpook National University, Daegu 41566, Korea Department of English Linguistics and Language Technology, Hankuk University of Foreign Studies, Seoul 02450, Korea; parkjs@hufs.ac.kr Department of Computer Science and Engineering, Sogang University, Seoul 04107, Korea; kimjihwan@sogang.ac.kr School of Electronics Engineering, Kyungpook National University, Daegu 41566, Korea * Correspondence: gjang@knu.ac.kr; Tel.: +82-53-950-5517 Featured Application: Automatic speech recognition; chatbot; voice-assisted control; multimodal man-machine interaction systems. Abstract: Speech recognition consists of converting input sound into a sequence of phonemes, then finding text for the input using language models. Therefore, phoneme classification performance is a critical factor for the successful implementation of a speech recognition system. However, correctly distinguishing phonemes with similar characteristics is still a challenging problem even for state-of- the-art classification methods, and the classification errors are hard to be recovered in the subsequent language processing steps. This paper proposes a hierarchical phoneme clustering method to exploit more suitable recognition models to different phonemes. The phonemes of the TIMIT database are carefully analyzed using a confusion matrix from a baseline speech recognition model. Using automatic phoneme clustering results, a set of phoneme classification models optimized for the generated phoneme groups is constructed and integrated into a hierarchical phoneme classification method. According to the results of a number of phoneme classification experiments, the proposed Citation: Oh, D.; Park, J.-S.; hierarchical phoneme group models improved performance over the baseline by 3%, 2.1%, 6.0%, and Kim, J.-H.; Jang, G.-J. Hierarchical Phoneme Classification for Improved 2.2% for fricative, affricate, stop, and nasal sounds, respectively. The average accuracy was 69.5% Speech Recognition. Appl. Sci. 2021, and 71.7% for the baseline and proposed hierarchical models, showing a 2.2% overall improvement. 11, 428. https://doi.org/10.3390/ app11010428 Keywords: speech recognition; phoneme classification; clustering; recurrent neural networks Received: 2 December 2020 Accepted: 30 December 2020 Published: 4 January 2021 1. Introduction These days, automatic speech recognition (ASR) performance has improved greatly by Publisher’s Note: MDPI stays neu- using deep neural networks [1–5]. Since speech is a time series, recurrent neural networks tral with regard to jurisdictional clai- (RNNs) are mostly adopted for speech recognition [6–8], but there have been several ms in published maps and institutio- attempts to apply a convolution neural network (CNN) as well [9,10]. In most ASR systems, nal affiliations. speech is first converted to a sequence of phonemes, which is then translated to a word sequence using lexical models and language models generate meaningful sentences from the word sequence. Therefore, in ASR, the correct prediction of phonemes is the most Copyright: © 2020 by the authors. Li- critical factor because the errors in the earlier stage cannot be recovered in the final text. censee MDPI, Basel, Switzerland. Phonetically, the phonemes are distinguished according to the variations in pronunci- This article is an open access article ation and vocalization. However, most speech recognition methods do not actively consider distributed under the terms and con- different characteristics of phonemes, such as end-to-end speech recognition methods [11–16]. ditions of the Creative Commons At- Phonemes with similar characteristics are often confused and hard to distinguish, and tribution (CC BY) license (https:// they degrade the overall phoneme classification performance. Choosing feature extrac- creativecommons.org/licenses/by/ tion methods and acoustic model types appropriately for confusing phonemes can help 4.0/). Appl. Sci. 2021, 11, 428. https://doi.org/10.3390/app11010428 https://www.mdpi.com/journal/applsci Appl. Sci. 2021, 11, 428 2 of 17 improve the final sentence recognition performance. We propose a novel method of apply- ing phoneme-specific acoustic models for automatic speech recognition by a hierarchical phoneme classification framework. The hierarchical phoneme classification is composed of a single, baseline phoneme classifier, clustering into similar groups and final result generation using retrained group- specific models. First, we analyze the result of a general phoneme classifier based on long short-term memory (LSTM) [17]. In the next stage, we proceed clustering the confusing phonemes using a confusion matrix to group phonemes with similar characteristics. Then the phoneme groups are analyzed to find the characteristics that are common inside and different across the groups. Consonants and vowels are distinguished well and the length of consonants is usually shorter. In the proposed method, various lengths for the analysis frame are compared and the best one is chosen according to the phoneme recognition accuracy. In addition, there are several pairs of different consonants that differ by being voiced or unvoiced. For example, ‘d’ and ‘t’ sounds in different words ‘dean’ and ‘teen’, respectively, or ‘b’ and ‘p’ sounds in ‘bad’ and ‘pad’. Those consonants can be distinguished by the existence of the glottal pulse that occurs at periodic time intervals [18–20], and we use autocorrelation functions to add the periodicity feature of the phoneme sound if the found phoneme falls into consonant categories. The proposed hierarchical framework uses group-specific models with different features for different phoneme groups, trained by separate training data for each phoneme group. The rest of the paper is organized as follows: Section 2 describes the phoneme cluster- ing process from the confusion matrix of the baseline phoneme classification model and shows the resultant phoneme groups by the clustering. Section 3 explains a hierarchical phoneme classification model that adds a phoneme group model to the baseline phoneme classification and evaluate its performance. In Section 4, the evaluation results of the proposed method on phoneme recognition for the TIMIT dataset are given. 2. Phoneme Clustering This section describes the basic definition of phonemes, particularly in English, the phoneme classes used in this paper, and the baseline phoneme classification method using RNN. Then the pairwise misclassification rates between all the pairs of the phonemes are obtained to build a confusion matrix, and phoneme clustering results are obtained by grouping phonemes so that the phonemes within the group should be more confused than the phonemes between the groups. The phoneme clustering results are used in building the proposed hierarchical classifier in Section 3. 2.1. Phonemes Phonemes are the set of units representing the distinctive sounds of a spoken language. In most languages, there are about 20 to 60 units. In this paper, we adopt ARPAbet [21], developed by the Advanced Research Projects Agency (ARPA), which exploits standard English alphabet symbols to represent IPA (International Phonetic Alphabet) symbols [22]. Table 1 shows the mapping from ARPAbet to IPA symbols, with 11 more symbols [23] that are used in the TIMIT dataset [24,25] for a more detailed phoneme analysis. The symbol [h#] is not an actual phoneme but a special indicator symbol for the beginning and ending of a sentence, so the number of additional TIMIT phonemes is 11. The phonemes are broadly classified into consonants and vowels [26], and they are shown in the first and second columns of Table 1. The number of vowels and consonants used in ARPAbet are 19 and 31, so the total number of phonemes in this paper is 61. The consonants and vowels are further divided into more detailed categories according to the method of vocalization [27,28]. The most common types of consonants are stops, fricatives, affricates, nasals, and semivowels [29]. Stops are generated by the constriction and release of air in the mouth (closing-compression-release) [30], so there exists a certain duration called “stop gap” when the air is compressed inside the mouth. The stop sounds usually have large energies in low frequencies, roughly below 200 Hz. Fricatives have a Appl. Sci. 2021, 11, 428 3 of 17 looser constriction in the mouth, which allows producing some amount of friction at the point of contact. Most of their acoustic energy occurs at higher frequencies. Fricatives are divided into sibilants versus non-sibilants [31]. Sibilants tend to be louder than non- sibilants, and turbulence noise is stronger in sibilants than non-sibilants. Nasals have the same constriction as plosives except that air is allowed to pass through the nose but not through the mouth [32]. Nasals show the abrupt loss of overall energy because the nose is less efficient than the mouth in radiating the energy to the outside. Nasals involve a voiced feature in the frequency domain and in addition, they have formant frequencies like vowels. Table 1. English phonetic symbols of the extended 2-letter ARPAbet used in the TIMIT dataset with mapping to IPA (International Phonetic Alphabet) symbols. The number of vowels is 19 and the number of consonants is 31. The begin/end marker is ignored in the phoneme recognition. There are 11 more special symbols used in the TIMIT dataset, so the total number of phonemes is 61. Vowels (19) Consonants (31) TIMIT Extension (11 + 1) ARPAbet IPA ARPAbet IPA ARPAbet IPA Symbol Description Devoiced [@] aa A b b q P ax-h ([@]) ae æ ch Ù r ô eng Syllabic [N] ah 2 d d s s hv Voiced [h] ao O dh D sh S bcl [b] closure aw aU dx R t t dcl [d] closure ax @ el l th T gcl [g] closure axr Ä em m v v kcl [k] closure ay aI en n w w pcl [p] closure eh E f f wh û tcl [t] closure er Ç g g y j pau Pause Epenthetic ey eI hh h z z epi silence Begin/end ih I jh à zh Z h# marker ix 1 k k iy i l l ow oU m m oy OI n n uh U ng N uw u nx  R ux 0 p p The set of 61 phonemes are mapped to a smaller set of 39 labels [33] are shown in Table 2. In the new phoneme classes with multiple phonemes (classes 2, 5, 6, 8, 14, 15, 19, 20, 21, 28, 38, and 39), similar phonemes are grouped together because they are known to be hard to distinguish with acoustic features. All the closure sounds, pause, and glottal plosive ([q], equivalent to IPA [P]) are all grouped together to mapped phoneme 39. They are relatively weak, are often accompanies by other sounds, and are very hard to distinguish by only sound signals without surrounding phonemes. 2.2. Baseline Phoneme Recognition with TIMIT Dataset TIMIT corpus is designed to provide speech data for studies of acoustic-phonetic and speech recognition systems [24,25]. Phonetic transcripts for all sentences are provided in the TIMIT corpus distribution. A set of 61 phoneme labels from ARPAbet symbols are used in the transcripts. We subsequently collapsed the set of 61 labels into a smaller set of 39 labels in Table 2 during the scoring of performance of phoneme recognition. The baseline monophone classifier is developed by a recurrent neural network (RNN) [34–36]. From the input speech signals, mel-frequency cepstral coefficients (MFCCs) [37,38] are extracted from the input speech, their first and second order derivatives are computed, Appl. Sci. 2021, 11, 428 4 of 17 and the resultant 39 dimensional vector is obtained at every 10 milliseconds analyzed by 25-millisecond-overlapping windows. In this feature extraction process, all phonemes are pre-processed identically. The input to the baseline RNN is the 39-dimensional MFCC feature vectors and the number of output nodes is 39 one-hot vectors according to the 39 labels of the phonemes. Table 2. Mapping 61 phonemes to 39 phonemes for phoneme recognition. The grouping is based on the linguistic similarities between phonemes [33]. No ARPAbet No ARPAbet No ARPAbet No ARPAbet 1 iy 11 oy 21 ng eng 31 b 2 ih ix 12 aw 22 v 32 p 3 eh 13 ow 23 f 33 d 4 ae 14 er axr 24 dh 34 dx ax ah 5 15 l el 25 th 35 t ax-h 6 uw ux 16 r 26 z 36 g 7 uh 17 w 27 s 37 k 8 ao aa 18 y 28 zh sh 38 hh hv bcl pcl 9 ey 19 m em 29 jh 39 dcl tcl gcl kcl q 10 ay 20 n en nx 30 ch epi pau 2.3. Confusion Matrix A confusion matrix is a table of comparison results for the predicted and actual values to analyze intra-class and inter-class performances. It is used to evaluate the performance of supervised learning, such as classification in machine learning [39]. After training the baseline phoneme classification model, on a test set we evaluate phoneme classification accuracies and create a confusion matrix. Figure 1 shows the obtained confusion matrix from baseline phoneme classification results for 39 mapped phonemes. The y-axis is the real labels of the given phonemes, and the x-axis is the predicted labels from the baseline RNN. Each element of the matrix, denoted by C(i, j) where i and j are row and column numbers, respectively, is the rate of predicting phoneme j when the actual one is i. According to the definition of the confusion matrix, the diagonal elements (i = j) are the correct classification rates for individual phonemes and all off-diagonal elements (i 6= j) are ratios of incorrect classification pairs. For example, the first column of the the third row, C(1, 3), is brighter than the nearby components with the of the exception diagonal ones. It means that the phoneme [aw] is misclassified a lot to the phoneme [ae]. Among off-diagonal elements, ([uh], [ax]), ([uh], [ih]), ([n], [m]), and ([z], [s]) are examples of phoneme pairs with high misclassification rates. 2.4. Phoneme Clustering Using Confusion Matrix We clustered phonemes using the similarities defined by the sum of the off-diagonal elements in the confusion matrix as shown in Figure 1. We use an agglomerative method based on bottom up clustering, which starts from a unique cluster and ties up a pair one at a time [40]. To define the distance between a phoneme pair, we use the confusion error rates, obtained by one of the off-diagonal elements of the confusion matrix. For a newly created phoneme cluster, the new error rate is computed by the average of the merged phonemes. The process stops until the desired number of clusters remain. The results of hierarchical clustering are usually presented in a dendrogram [41]. As shown in Figure 2, the dendrogram is a tree diagram representing clusters obtained from hierarchical clustering. The y-axis shows phoneme labels and the phonemes are hierarchically paired with the height of the pair connection being the misclassification rate between the phonemes Appl. Sci. 2021, 11, 428 5 of 17 in the corresponding pair. The misclassification rate range from 0.0 to 1.0 along the x-axis. This method is called hierarchical clustering because paired phonemes are considered as a single phoneme after the pair is merged and all phonemes finally converge to a single cluster. The goal of the the proposed method is to improve the performance of phoneme recognition. Therefore, for the confusing phonemes, other classification method that is different from the baseline phoneme classifier is applied to further distinguish the phonemes in the same cluster. ���������������� �� �� �� �� �� ��� �� �� �� �� �� �� �� ��� �� �� �� �� ����� ��� �� �� �� ��� �� �� �� �� ��� � � � � � � � � � � � � � � �� �� �� �� � � �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� ���������� Figure 1. Confusion matrix from the results of the baseline phoneme classification. Each row of the matrix represents the instances in a truth and each column represents the instances in a predicted class. th dh hh dx cl ng jh ch zh uh ax ih eh uw aw ae er iy ey ow ay ao oy 0.0 0.2 0.4 0.6 0.8 1.0 Figure 2. Dendrogram of clustering. The x-axis shows a similarity of the clustered nodes. The higher the x-value, the more similar they are to each other. The y-axis is the phoneme label. This dendrogram shows that the consonants and vowels are well distinguished. Appl. Sci. 2021, 11, 428 6 of 17 In Figure 2, if we backtrack the clustering path from the finally merged cluster at the right end, all the phonemes are broadly divided into two clusters. The largest cluster in the upper side is a set of phonemes from [th] to [zh]. The lower largest cluster consists of phonemes from [uh] to [oy], most of which are vowel sounds. However, [l] and [r] are consonants and more specifically, belong to voiced alveolar sounds. The phonemes [w] and [y] are semivowels, belonging to voiced labio-velar and voiced palatal sounds, respectively. These sounds are between vowels and consonants, so we split the lower cluster into “Vowels” and “Mixed”, as shown in the first and second rows of Table 3. Among vowels, the phonetic classification is somewhat grouped according to the pronunciation position similarly to the conventional phonetic classification, but is not accurately distinguished. This is because the pronunciation of vowels vary according to the adjacent phonemes. All of the phonemes in the upper largest cluster are consonants. The individual consonants have different characteristics, so we divided the consonant set into four major groups. The lower four rows of Table 3 are the phoneme groups based on the dendrogram in Figure 2 using a 0.6 cut-off with some additional adjustments. The first group is a set of nasal sounds, [m], [n], and [ng]. The second one is a group of fricatives, sounds with air friction with longer pronunciation time. Voiced sounds ([z] [zh] [v] [dh]) and their corresponding unvoiced sounds ([s] [f] [th]) are grouped together. The voiceless glottal fricative sound, [hh], also belongs to this group. The third is the affricates group ([jh] [ch]), weak sounds with friction. The stop sounds such as [t], [d], [p], and [b] are closely clustered. The clustering result is similar to the basic phonetic categories. Our resultant clusters are analogous to the conventional linguistic categories, so we can confirm that it provides reasonable results. Table 3. Phoneme groups obtained by agglomerative hierarchical clustering on the TIMIT train- ing dataset. Phonemes of the Group Group (Written in ARPAbet) iy ih eh ae ax uw uh Vowels ao ey ay oy aw ow er Mixed l r w y Nasals m n ng Fricatives s z zh f v th dh hh Consonants Affricates jh ch Stops b d g k p t dx To confirm that the phoneme categories are well clustered, we drew a confusion matrix with the new phoneme groups in Table 3, and it is shown in Figure 3a. The phonemes are reordered so that all the phonemes in the same group are located adjacently. The red-colored rectangles indicate phoneme groups, and Figure 3b is the table below the figure is a confusion matrix with the phoneme groups as classification targets. The diagonal ones represent that the accuracy of the group labeling is very high, which implies that the different groups are well distinguished. In other words, the phonemes within the same group are similar and confusing, so we focus on improving the recognition performance of these phonemes inside the groups. Appl. Sci. 2021, 11, 428 7 of 17 (a) Phoneme grouping in the confusion matrix. vowel 0.94 0.02 0.02 0.01 0.00 0.01 mixed 0.02 0.92 0.01 0.01 0.00 0.04 Real nasal 0.07 0.09 0.78 0.01 0.00 0.04 fric 0.01 0.04 0.01 0.88 0.01 0.06 affr 0.00 0.01 0.00 0.07 0.80 0.12 stop 0.05 0.05 0.01 0.04 0.01 0.85 vowel mixed nasal fric affr stop Prediction (b) Confusion matrix of the phoneme groups. Figure 3. (a) The confusion matrix rearranged according to the phoneme clustering result. Each row of the matrix represents the instances in a truth while each column represents the instances in a predicted class. (b) Confusion matrix of phoneme groups obtained by phoneme clustering. 3. Hierarchical Phoneme Classification This section analyzes the characteristics of clustered phoneme groups. We explain the similarities and differences between phoneme groups, and the detailed implementation of group-specific models. By combining the group-specific phoneme classification models, a hierarchical phoneme classification framework is proposed. All individual group models are implemented by deep neural networks. The neural network automatically finds and recognizes the characteristics of speech signals. For sequential inputs such as speech signals, modeling temporal patterns is necessary. A recurrent neural network (RNN) is a type of neural network that can effectively learn temporal data. Therefore, phoneme classification model in this paper is built using RNN. More specifically, bi-directional long short-term memory (BLSTM) is adopted [8,42,43]. 3.1. Overall Architecture We propose a hierarchical phoneme classification model to improve the performance of phoneme classification. Figure 4 shows the overall model architecture. First, phoneme is predicted through the baseline phoneme classification model in which a whole phoneme is learned by using the same feature extraction method without distinguished characteristics. The baseline model was constructed using long short-term memory (LSTM) [17]. LSTM is Real Appl. Sci. 2021, 11, 428 8 of 17 an RNN cell that can avoid the problem of vanishing gradient using three gates of input, forget, and output. A typical LSTM uses a unidirectional structure that learns only by check- ing information from previous contexts. In the speech recognition, a bi-directional structure that learns both past and future information shows good performance [42–44]. Therefore, we implemented the model with the bi-directional LSTM (BLSTM) as a baseline system for phoneme recognition [45]. Three BLSTM layers were stacked and a fully-connected layer is added. At the output, the phoneme is classified using the softmax layer. Vowels vowel model Mixed: [l] [r] [w] [y] mixed model nasals feature extraction model by group Predicted baseline Phoneme feature different window Phoneme phoneme clas- wave extraction size fricatives sification Label (MFCC_gws) model affricates Consonant model autocorrelation stops model Figure 4. Overall model structure of the proposed hierarchical classification. Using the baseline phoneme classifier, phoneme-dependent features are extracted and different phoneme group models are applied to obtain refined phoneme classification results. The learning target of the predicted phoneme for the baseline model is the index of the phoneme group in Table 3. Then, the speech feature is re-extracted using the optimal window size for each phoneme group. In addition, autocorrelation is used together to dis- tinguish voiced and unvoiced phonemes. The extracted speech features are used as inputs to each phoneme group model. The phoneme group model finally predicts the phoneme. 3.2. Vowels and Mixed Phoneme Classification Two models for phoneme groups, ‘Vowel’ and ‘Mixed’ are trained by using group- specific training data. The ‘Vowel’ has 14 different labels as listed in Table 3, which is reduced from 39 phoneme labels of the baseline classifier. Out of the TIMIT training dataset, audio segments labeled by iy, ih, . . ., er are extracted and used in training ‘Vowel model’. To re-train ‘Mixed model’, training subsets labeled by l, r, w, and er are used. The model architecture is shown in Figure 5. Predicted Phoneme BLSTM FC MFCC Phonemes wave layers layer Figure 5. Phoneme classification model architecture for the baseline, as well as Vowel and Mixed groups. 3.3. Varying Analysis Window Sizes for Consonants Consonants are short and instantaneous, and vowels are longer and more consis- tent [46]. It would be better to shorten the analysis window size in feature extraction to more accurately distinguish short consonants. Therefore, we empirically decided appro- priate analysis window lengths differently for individual phoneme groups. The baseline phoneme recognition model extracts features using MFCC with the same conditions such Appl. Sci. 2021, 11, 428 9 of 17 as same window and shift sizes for all phoneme. In this process, a 16 kHz sampling rate and 25 ms hamming window is used. The window shift was set to 10 ms. However, conso- nants are very short and strong compared to vowels. Since the properties of phonemes are different, we compared the performance of phoneme recognition by extracting features while changing window and shift sizes for each group. The window sizes are 5 ms, 10 ms, 15 ms, 20 ms, 25 ms, and 30 ms. ‘MFCC_gws’ in Figure 4 is the MFCC features with varying window sizes. 3.4. Voiced and Unvoiced Consonants Classification Several pairs of different consonants are distinguished by being voiced or unvoiced. For voiced sounds, the glottis is closed or nearly closed, whereas for voiceless sounds it is wide open, with the distance between the folds actually amounting to only a fraction of a millimeter [18]. The voiced sounds carry periodic pulses generated by the glottis vibration, while almost no periodicity but turbulent noise is observed in the unvoiced sounds. Phonemes [b], [p], [d], and [t], all belong to the stops group in the phonetic classification. However [b], [d] are voiced phonemes and [p], [t] are unvoiced ones. Phonemes [z] and [s] are both fricatives, but [z] is voiced and [s] unvoiced. These pairs of examples are usually hard to distinguish because they are generated at the almost same place in the oral cavity area. Similarly, [p] and [b] are produced at the same place in the mouth and the only difference is being voiced or unvoiced. If we add a feature that helps distinguish voiced and unvoiced phonemes especially when recognizing consonants and create a model that can classify them, the performance may be improved. Therefore, we use autocorrelation functions to add the periodicity feature of the phoneme sound if the found phoneme falls into consonant categories Voiced and unvoiced sounds can be distinguished by pitch, or equivalently funda- mental periodicity values [47]. Autocorrelation measures the similarity function of the time lag with itself in the signal [19,20]. Due to the vibration of the vocal cords, the voiced sound is periodic and this vibration interval is pitch. Autocorrelation is a unique case of cross-correlation [48,49]. The cross-correlation measures the similarity of the two discrete signals, x[n] and y[n], defined by [50]: x[n + t]y[n] , (1) n=¥ where t is an integer time lag that takes values from f0,1,2, . . . ,g. If two discrete signals are identical, this process is defined as autocorrelation, which is given by: ac(x, t) = x[n + l]x[n] . (2) n=¥ Algorithm 1 is the pseudo code for the autocorrelation computation. The array x is the speech signal, N is the size of signal, and the autocorrelation function is denoted by ac. Figure 6a shows the waveform of a voiced consonant phoneme [d] and its autocorrelation computed by Algorithm 1. Likewise, Figure 6b is for an unvoiced consonant phoneme [t]. The autocorrelation of phoneme [d] shows nonzero values, while that of phoneme [t] has almost zero, meaning that there is no significant pitch. From these results, autocorrelation can help in distinguishing voiced and unvoiced phonemes. Appl. Sci. 2021, 11, 428 10 of 17 Algorithm 1 Autocorrelation H] ac array of autocorrelation N length of signal x for i 0 to N do sum 0 for j 0 to N 1 do sum sum + (x[j + 1] x[j]) end for ac[i] sum/N end for 3.5. Consonant Group Model Architectures As described in the previous section, we focus on making a better decision on con- sonants rather than on vowels. Consonants show various types of energy distributions in phoneme groups. Therefore, we implement appropriate models for individual conso- nant phoneme groups and verify their performance. MFCC_gws is a feature extraction method with group different window sizes, as well as autocorrelation features added to help distinguish voiced phonemes and unvoiced phonemes within the consonant group. By combining the two features [51], three phoneme group models are created. Figure 7 shows three different model configurations. Figure 7a simply concatenates MFCC_gws and the autocorrelation vector at the input unit and passes them to a BLSTM layer. A fully connected (FC) layer generates softmax outputs as many as the number phoneme to generate predictions for phoneme labels. Figure 7b has individual BLSTM layers for MFCC_gws and the autocorrelation vector, and concatenates the outputs to pass them through an FC layer. The advantage of output concatenation over input concatenation is that the mismatches of the different kinds of features, such as scale differences can be automatically adapted to each other. The concatenation at the BLSTM output layers is expected to be more reliable than Figure 7a. Figure 7c adds another BLSTM layer to the concatenated output of Figure 7b. These three configurations are used to construct the hierarchical phoneme classifier in Figure 4. 2 − 00 2 − 00 0 50 100 150 200 250 300 0 50 100 150 200 250 300 350 times times 1 − 000 1 − 000 2 − 000 0 50 100 150 200 250 300 0 50 100 150 200 250 300 350 times times (a) Voiced phoneme [d]. (b) Unvoiced phoneme [t]. Figure 6. Waveforms and autocorrelation functions of a voiced sound (phoneme [d]) and a unvoiced sound (phoneme [t]). Upper floats are waveforms and lower floats are computed autocorrelation functions. Relative amplitude amplitude Relative amplitude amplitude Appl. Sci. 2021, 11, 428 11 of 17 autocorrelation Predicted Phoneme BLSTM FC concat Phoneme wave layers layer MFCC_gws (a) Input concatenation. BLSTM autocorrelation outputs layers Predicted Phoneme FC concat Phoneme wave layer BLSTM MFCC_gws outputs layers (b) Concatenated bi-directional long short-term memory (BLSTM) and fully connected network (FCN). BLSTM autocorrelation outputs layers Predicted Phoneme BLSTM FC concat Phoneme wave layers layer BLSTM MFCC_gws outputs layers (c) Concatenated BLSTM, and BLSTM and FCN. Figure 7. Various types of the consonant group models with additional autocorrelation features. (a) input features are concatenated. (b) BLSTM outputs are cocatenated. (c) another BLSTM layer is added for the concatenated BLSTM outputs. 4. Experimental Results This section shows the analysis results of the phoneme group characteristics described in Section 2 and also presents comparison results of the performance of the phoneme group model with that of the baseline model. 4.1. TIMIT Database We have carried out speech recognition experiments on the TIMIT corpus [25], which is designed to provide speech data for studies of acoustic-phonetic and speech recognition systems. The speech data was recorded at Texas Instruments (TI), transcribed at the Massachusetts Institute of Technology (MIT), and prepared for CD-ROM production by the National Institute of Standards and Technology (NIST). The corpus includes the speech waveforms with phonetic transcriptions. TIMIT corpus has 6300 sentences, spoken by Appl. Sci. 2021, 11, 428 12 of 17 630 speakers, who spoke exactly 10 sentences each. There are three types of sentences. Dialect sentences (SA) were recorded from speakers with dialectal diversity. Phonetically- compact (SX) and phonetically-diverse sentences (SI) were created to model different situations of phoneme distributions. Phonetic transcripts for all sentences are provided in the TIMIT corpus distribution. The full phoneme transcript of an example sentence is shown in Table 4. All the numbers are in samples, so the length of the audio file in the example is 46,797/16,000 kHz = 2.924 s. Table 4. An example of the phonetic transcripts in the TIMIT corpus. The audio file is “dr1/rcjf0/sa1.wav” in the provided CD-ROM, and its sentence transcript is “sa1.txt” in the same directory. The phonetic transcript file is “sa1.phn”, and full timings of phonemes are provided as well. File Begin End Text/Phoneme She had your dark suit in greasy sa.txt 0 46797 wash water all year. 0 3050 h# 3050 4559 sh 4559 5723 ix sa1.phn 5723 6642 hv 6642 8772 eh 8772 9190 dcl 9190 10,337 jh . . . . . . . . . 40,313 42,059 y 42,059 43,479 ih 43,479 44,586 axr 44,586 46,720 h# The whole corpus is divided into non-overlapping training and testing sets. We use core test set [25] in an evaluating performance of phoneme recognition. The core test data has a total of 24 speakers, more specifically two male and one female from each of the eight dialect regions. The number of phoneme tokens are 177,080 in the training set and 9098 in the testing set. The number of tokens for each of the five phoneme groups is shown in Table 5. Table 5. Phoneme token counts of the TIMIT train and core test sets, grouped by phoneme classes. Phoneme Fricatives Affricates Stops Nasals Vowels/Mixed Groups Train set 21,424 2031 22,281 14,157 75,257 Core test set 1114 95 1114 736 3955 4.2. MFCC vs. Log Filter Bank Energy We compared the performance of two commonly-used features in speech recognition. The first one is mel-frequency cepstral coefficients (MFCCs) and the other one is log filterbank energies. Figure 8 shows the performance differences of four phoneme groups according to the change of window sizes. Blue lines are MFCC results and orange lines are the results of log filter bank energy features [37,38,52]. In most cases, blue lines are below orange lines, showing less phoneme recognition errors. Based on these results, we chose MFCC in the phoneme group models and the overall model. Appl. Sci. 2021, 11, 428 13 of 17 Fricatives 0.35 MFCC logfbank 0.30 0.25 0.20 0.15 5 10 15 20 25 30 Affricates MFCC 0.27 logfbank 0.26 0.25 0.24 0.23 5 10 15 20 25 30 Window size [ms] (a) Fricatives and Affricates. Stops 0.40 MFCC logfbank 0.35 0.30 0.25 0.20 5 10 15 20 25 30 Nasals MFCC 0.30 logfbank 0.28 0.26 5 10 15 20 25 30 Window size [ms] (b) Stops and Nasals. Figure 8. The error rates using mel-frequency cepstral coefficients (MFCCs) and log filter bank energy according to change window size. (a) shows fricatives and afficates. (b) shows stops and nasals. In all phoneme groups, the error rate of MFCC is lower than that of log filter bank energy. 4.3. Various Window Sizes In order to find the optimal window sizes for individual phoneme groups, we tried various window sizes in the MFCC feature extraction. Table 6 shows performance varia- tions by choosing different window sizes with six phoneme group models. For fricatives and stops, 10 ms window size was the best, both 15 ms and 25 ms were the best for af- fricates ([jh] and [ch]) and 20 ms was the best for nasal. This is because fricative and stop sounds are very short consonants, and nasal sounds are voiced and close to vowel sounds. Generally, consonant models showed lower phoneme error rates with smaller window sizes because the actual vocalization length is shorter for consonants, so it is possible to obtain more information by using a small window length. Surprisingly, for vowel/semivowel and closes, window analysis size did not affect performance. Unlike consonants, they have very stable characteristics over time, so a change in analysis sizes error rate error rate error rate error rate Appl. Sci. 2021, 11, 428 14 of 17 made no difference. We used the best window sizes for each of the phoneme group models to maximize the performance of the proposed hierarchical phoneme classification model. 4.4. Phoneme Group Model Training The baseline model uses MFCC features only. Clustering result showed that voiced and unvoiced phonemes are hard to distinguish. To overcome this problem, autocorrelation feature is added in phoneme group models because it shows very different values for voiced and unvoiced speech signals. Table 7 shows the performance of phoneme group models. This result shows that using autocorrelation is better than the MFCC alone. In addition, the performance of the three models was different for each phoneme group. Since the phoneme group has different characteristics, the optimal model structure is also different. Based on these results, we constructed a hierarchical model using phoneme group models. 4.5. Performance of the Hierarchical Classification We implemented a hierarchical phoneme classification model that combines all the phoneme group models with the optimal window sizes for the corresponding phoneme groups and evaluated the performance. We used the Adam optimizer, which adaptively adjusts the learning rate of the backpropagation learning algorithm for the artificial neural networks [53]. The initial learning rate of the Adam optimizer is set to be 10 , and gradually lowered as the learning proceeds. Batch size 512 is used during training of each phoneme group model. Dropout [54] with a probability of 0.2 is added across the BLSTM layers. Table 8 represents the performance of the overall model. Compared with the baseline, models with different window sizes for phoneme groups had an improved performance. In addition, the overall model using the phoneme group model combining MFCC and autocorrelation showed better performance. Table 6. The results of overall model using MFCC with varying window sizes. Analysis Window Sizes Baseline 5 ms 10 ms 15 ms 20 ms 25 ms 30 ms Vowels 0.700 - - - - - - Mixed 0.919 - - - - - - Nasals 0.568 0.565 0.553 0.564 0.579 0.557 0.568 Fricatives 0.724 0.710 0.739 0.738 0.737 0.735 0.736 Affcicates 0.632 0.632 0.642 0.653 0.590 0.653 0.642 Stops 0.628 0.657 0.682 0.651 0.647 0.626 0.620 Group 0.695 0.697 0.706 0.704 0.695 0.698 0.698 avg Integrated 0.730 0.732 0.738 0.735 0.735 0.731 0.731 Table 7. The results that different window sizes in MFCC and MFCC with autocorrelation and a three feature ensemble model. MFCC/Autocorrelation Combinations Phoneme Group Specific Concat BLSTM Concat BLSTM Input Concat + BLSTM + Groups Window Sizes + FCN FCN Nasals 0.742 0.757 0.742 0.757 Fricatives 0.844 0.847 0.834 0.827 Affricates 0.768 0.737 0.779 0.747 Stops 0.796 0.781 0.806 0.778 Appl. Sci. 2021, 11, 428 15 of 17 Table 8. The results that baseline and integration phoneme group model. Overall Model Phoneme Baseline Group MFCC and Groups Phoneme Specific Autocorrealation Classification Window Size Combination Model Vowels 0.700 0.700 0.700 Mixed 0.919 0.919 0.919 Nasals 0.568 0.579 0.590 Fricatives 0.724 0.739 0.754 Affcicates 0.632 0.653 0.653 Stops 0.628 0.682 0.688 Group avg 0.695 0.712 0.717 Integrated 0.730 0.740 0.743 5. Discussion In this paper, we proposed a hierarchical speech recognition model based on phoneme clustering. The proposed model first classifies the entire phoneme by a baseline phoneme classifier without considering phoneme characteristics. Then the retrained phoneme group models are used using the initial phoneme classification results. A confusion matrix of baseline phoneme classification was used to group each phoneme with similar charac- teristics. The confusion matrix shows which phonemes are not well distinguished. An analysis of phoneme characteristics in the TIMIT database revealed the differences of the phonemes in terms of lengths and autocorrelation values, so they were used in constructing phoneme group-specific models. We extracted features using different window size and window shift optimized for each phoneme group, and used autocorrelation as a feature as well as MFCC to improve the performance of the phoneme group models. The results of the experiment of the four consonant phoneme group models showed that the phoneme groups have different characteristics, and thus, the suitable models are different. The hier- archical phoneme classification model consisting of the baseline model and the phoneme group model was evaluated. The result shows that the phoneme specific feature extraction method and classification model have an improved performance of phoneme classification and speech recognition. One drawback of the proposed hierarchical classification method is that it does not improve the vowel classification performance because the proposed autocorrelation features and varying window sizes only help consonant classification. The MFCC is well suited to vowel classification and almost optimal classification performance is achieved, so very little improvement is expected. Author Contributions: Conceptualization, J.-S.P., J.-H.K., and G.-J.J.; methodology, D.O.; software, D.O.; validation, D.O. and G.-J.J.; formal analysis, J.-H.K. and J.-S.P.; investigation, J.-H.K.; resources, G.-J.J. and J.-H.K.; data curation, D.O.; writing–original draft preparation, D.O. and G.-J.J.; writing– review and editing, J.-S.P. and J.-H.K.; visualization, D.O.; supervision, G.-J.J.; project administration, G.-J.J.; funding acquisition, G.-J.J. All authors have read and agreed to the published version of the manuscript. Funding: This work was supported by Institute for Information and Communications Technology Promotion (IITP) grant funded by the Korean government (MSIT) (No. R7124-16-0004, Development of Intelligent Interaction Technology Based on Context Awareness and Human Intention Understand- ing), by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. NRF-2017M3C1B6071400), and the BK21 Plus project funded by the Ministry of Educa- tion, Korea (21A20131600011). Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Appl. Sci. 2021, 11, 428 16 of 17 Data Availability Statement: The TIMIT dataset can be obtained from the Linguistic Data Consor- tium (LDC), https://www.ldc.upenn.edu. Conflicts of Interest: The authors declare no conflict of interest. Abbreviations The following abbreviations are used in this manuscript: ARPA Advanced Research Projects Agency ASR Automatic Speech Recognition BLSTM Bi-directional Long Short-Term Memory CNN Convolutional Neural Network IPA International Phonetic Alphabet LSTM Long Short-Term Memory MFCC Mel-Frequency Cepstral Coefficient NIST National Institute of Standards and Technology RNN Recurrent Neural Network References 1. Hinton, G.; Deng, L.; Yu, D.; Dahl, G.E.; Mohamed, A.R.; Jaitly, N.; Senior, A.; Vanhoucke, V.; Nguyen, P.; Sainath, T.N.; et al. Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups. IEEE Signal Process. Mag. 2012, 29, 82–97. [CrossRef] 2. Yu, D.; Li, J. Recent progresses in deep learning based acoustic models. IEEE/CAA J. Autom. Sin. 2017, 4, 396–409. 3. Xiong, W.; Wu, L.; Alleva, F.; Droppo, J.; Huang, X.; Stolcke, A. The Microsoft 2017 conversational speech recognition system. In Proceedings of the 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Calgary, AB, Canada, 15–20 April 2018; pp. 5934–5938. 4. Caterini, A.L.; Chang, D.E. Deep Neural Networks in a Mathematical Framework; Springer: New York, NY, USA, 2018. [CrossRef] 5. Taylor, S.; Kim, T.; Yue, Y.; Mahler, M.; Krahe, J.; Rodriguez, A.G.; Hodgins, J.K.; Matthews, I.A. A deep learning approach for generalized speech animation. ACM Trans. Gr. 2017, 36, 1–11. [CrossRef] 6. Graves, A. Sequence Transduction with Recurrent Neural Networks. arXiv 2012, arXiv:1211.3711. 7. Graves, A.; Mohamed, A.R.; Hinton, G. Speech recognition with deep recurrent neural networks. In Proceedings of the 2013 IEEE International Conference on Acoustics, Speech and Signal Processing, Vancouver, BC, Canada, 26–31 May 2013; pp. 6645–6649. 8. Zeyer, A.; Doetsch, P.; Voigtlaender, P.; Schlüter, R.; Ney, H. A Comprehensive Study of Deep Bidirectional LSTM RNNs for Acoustic Modeling in Speech Recognition. In Proceedings of the 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), New Orleans, LA, USA, 5–9 May 2017; pp. 2462–2466. 9. Abdel-Hamid, O.; Mohamed, A.R.; Jiang, H.; Deng, L.; Penn, G.; Yu, D. Convolutional neural networks for speech recognition. IEEE/ACM Trans. Audio Speech Lang. Process. 2014, 22, 1533–1545. [CrossRef] 10. Tóth, L. Phone recognition with hierarchical convolutional deep maxout networks. EURASIP J. Audio Speech Music Process. 2015, 2015, 25. [CrossRef] 11. Hannun, A.; Case, C.; Casper, J.; Catanzaro, B.; Diamos, G.; Elsen, E.; Prenger, R.; Satheesh, S.; Sengupta, S.; Coates, A.; et al. Deep speech: Scaling up end-to-end speech recognition. arXiv 2014, arXiv:1412.5567. 12. Miao, Y.; Gowayyed, M.; Metze, F. EESEN: End-to-end speech recognition using deep RNN models and WFST-based decoding. In Proceedings of the 2015 IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU), Scottsdale, AZ, USA, 13–17 December 2015; pp. 167–174. 13. Bahdanau, D.; Chorowski, J.; Serdyuk, D.; Brakel, P.; Bengio, Y. End-to-end attention-based large vocabulary speech recognition. In Proceedings of the 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Shanghai, China, 20–25 March 2016; pp. 4945–4949. 14. Zhang, Y.; Pezeshki, M.; Brakel, P.; Zhang, S.; Bengio, C.L.Y.; Courville, A. Towards end-to-end speech recognition with deep convolutional neural networks. arXiv 2017, arXiv:1701.02720. 15. Battenberg, E.; Chen, J.; Child, R.; Coates, A.; Gaur, Y.; Li, Y.; Liu, H.; Satheesh, S.; Seetapun, D.; Sriram, A.; et al. Exploring Neural Transducers for End-to-End Speech Recognition. CoRR 2017, abs/1707.07413. 16. Chiu, C.C.; Sainath, T.N.; Wu, Y.; Prabhavalkar, R.; Nguyen, P.; Chen, Z.; Kannan, A.; Weiss, R.J.; Rao, K.; Gonina, E.; et al. State-of-the-art speech recognition with sequence-to-sequence models. In Proceedings of the 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Calgary, AB, Canada, 15–20 April 2018; pp. 4774–4778. 17. Hochreiter, S.; Schmidhuber, J. Long short-term memory. Neural Comput. 1997, 9, 1735–1780. [PubMed] 18. Hayes, B. Introductory Phonology; Malden, M.A., Ed.; Wiley-Blackwell: Oxford, UK, 2009. 19. Rabiner, L. On the use of autocorrelation analysis for pitch detection. IEEE Trans. Acoust. Speech Signal Process. 1977, 25, 24–33. [CrossRef] Appl. Sci. 2021, 11, 428 17 of 17 20. Hernández, M. A tutorial to extract the pitch in speech signals using autocorrelation. Open J. Technol. Eng. Discip. (OJTED) 2016, 2, 1–11. 21. Klautau, A. ARPABET and the TIMIT Alphabet. 2001. 22. International Youth Association TIP. Reproduction of The International Phonetic Alphabet; Cambridge University Press: Lon- don, UK, 2005. 23. Gold, B.; Morgan, N.; Ellis, D. Speech and Audio Signal Processing: Processing and Perception of Speech and Music, 2nd ed.; Wiley: Hoboken, NJ, USA, 2011. 24. Linguistic Data Consortium. Table of All the Phonemic and Phonetic Symbols Used in the Timit Lexicon; Linguistic Data Consortium: Philadelphia, PA, USA, 1990. 25. Garofolo, J.; Lamel, L.; Fisher, W.; Fiscus, J.; Pallett, D. DARPA TIMIT acoustic-phonetic continous speech corpus CD-ROM. NIST speech disc 1-1.1. NASA STI/Recon Tech. Rep. N 1993, 93, 1–94. 26. Ladefoged, P.; Disner, S.F. Vowels and Consonants; John Wiley & Sons: Hoboken, NJ, USA, 2012. 27. Ades, A.E. Vowels, consonants, speech, and nonspeech. Psychol. Rev. 1977, 84, 524. [CrossRef] 28. Stevens, K.N. Acoustic Phonetics; MIT Press: Cambridge, MA, USA, 2000; Volume 30, 29. Miller, G.A.; Nicely, P.E. An analysis of perceptual confusions among some English consonants. J. Acoust. Soc. Am. 1955, 27, 338–352. [CrossRef] 30. Halle, M.; Hughes, G.W.; Radley, J.P. Acoustic properties of stop consonants. J. Acoust. Soc. Am. 1957, 29, 107–116. [CrossRef] 31. Shadle, C.H. The acoustics of fricative consonants. Tech. Rep. 1985, 1–200. [CrossRef] 32. Fujimura, O. Analysis of nasal consonants. J. Acoust. Soc. Am. 1962, 34, 1865–1875. [CrossRef] 33. Halberstadt, A.K. Heterogeneous Acoustic Measurements and Multiple Classifiers for Speech Recognition; Massachusetts Institute of Technology: Cambridge, MA, USA, 1999. 34. Evangelopoulos, G.N. Efficient Hardware Mapping of Long Short-Term Memory Neural Networks for Automatic Speech Recognition. Master ’s Thesis, KU Leuven, Belgium, 2016. 35. Ek, J.V.; Michálek, J.; Psutka, J. Recurrent DNNs and its Ensembles on the TIMIT Phone Recognition Task. arXiv 2018, arXiv:1806.07186. 36. Michaálek, J.; Ek, J.V. A Survey of Recent DNN Architectures on the TIMIT Phone Recognition Task. arXiv 2018, arXiv:1806.07974. 37. Han, W.; Chan, C.F.; Choy, C.S.; Pun, K.P. An efficient MFCC extraction method in speech recognition. In Proceedings of the 2006 IEEE International Symposium on Circuits and Systems, Island of Kos, Greece, 21–24 May 2006; p. 4. 38. Muda, L.; Begam, M.; Elamvazuthi, I. Voice recognition algorithms using mel frequency cepstral coefficient (MFCC) and dynamic time warping (DTW) techniques. arXiv 2010, arXiv:1003.4083. 39. Davis, J.; Goadrich, M. The relationship between Precision-Recall and ROC curves. In Proceedings of the 23rd International Conference on Machine Learning, Pittsburgh, PA, USA, 25–29 June 2006; pp. 233–240. 40. Corpet, F. Multiple sequence alignment with hierarchical clustering. Nucl. Acids Res. 1988, 16, 10881–10890. [CrossRef] 41. Wilks, D.S. Cluster analysis. In International Geophysics; Elsevier: Amsterdam, The Netherlands, 2011; Volume 100, pp. 603–616. 42. Schuster, M.; Paliwal, K.K. Bidirectional recurrent neural networks. IEEE Trans. Signal Process. 1997, 45, 2673–2681. [CrossRef] 43. Graves, A.; Schmidhuber, J. Framewise phoneme classification with bidirectional LSTM networks. In Proceedings of the 2005 IEEE International Joint Conference on Neural Networks, Montreal, QC, Canada, 31 July–4 August 2005; Volume 4, pp. 2047–2052. 44. Hannun, A.Y.; Maas, A.L.; Jurafsky, D.; Ng, A.Y. First-pass large vocabulary continuous speech recognition using bi-directional recurrent DNNs. arXiv 2014, arXiv:1408.2873. 45. Graves, A.; Fernández, S.; Schmidhuber, J. Bidirectional LSTM networks for improved phoneme classification and recognition. In International Conference on Artificial Neural Networks; Springer: New York, NY, USA, 2005; pp. 799–804. 46. Caramazza, A.; Chialant, D.; Capasso, R.; Miceli, G. Separable processing of consonants and vowels. Nature 2000, 403, 428. [CrossRef] [PubMed] 47. Rabiner, L.R.; Schafer, R.W. Digital Processing of Speech Signals; Pearson Publishing: NewYork, NY, USA, 1978; Volume 100. 48. Chen, C.H. Signal Processing Handbook; CRC Press: Boca Raton, FL, USA, 1988; Volume 51. 49. Furui, S.; Sondhi, M.M. Advances in Speech Signal Processing; Marcel Dekker, Inc.: New York, NY, 1991. 50. Proakis, J.G. Digital Signal Processing: Principles Algorithms and Applications; Pearson Education India Services Pvt.: Chennai, India, 51. Ngiam, J.; Khosla, A.; Kim, M.; Nam, J.; Lee, H.; Ng, A.Y. Multimodal deep learning. In Proceedings of the 28th International Conference On Machine Learning (ICML-11), Bellevue, WA, USA, 28 June–2 July 2011; pp. 689–696. 52. Nadeu, C.; Hernando, J.; Gorricho, M. On the decorrelation of filter-bank energies in speech recognition. In Proceedings of the Fourth European Conference on Speech Communication and Technology, Madrid, Spain, 18–21 September 1995. 53. Kingma, D.P.; Ba, J. Adam: A method for stochastic optimization. arXiv 2014, arXiv:1412.6980. 54. Srivastava, N.; Hinton, G.; Krizhevsky, A.; Sutskever, I.; Salakhutdinov, R. Dropout: A simple way to prevent neural networks from overfitting. J. Mach. Learn. Res. 2014, 15, 1929–1958. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Applied Sciences Multidisciplinary Digital Publishing Institute

Hierarchical Phoneme Classification for Improved Speech Recognition

Loading next page...
 
/lp/multidisciplinary-digital-publishing-institute/hierarchical-phoneme-classification-for-improved-speech-recognition-W0Vk8HL46q
Publisher
Multidisciplinary Digital Publishing Institute
Copyright
© 1996-2021 MDPI (Basel, Switzerland) unless otherwise stated Disclaimer The statements, opinions and data contained in the journals are solely those of the individual authors and contributors and not of the publisher and the editor(s). MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Terms and Conditions Privacy Policy
ISSN
2076-3417
DOI
10.3390/app11010428
Publisher site
See Article on Publisher Site

Abstract

applied sciences Article Hierarchical Phoneme Classification for Improved Speech Recognition 1,2 3 4 2,5, Donghoon Oh , Jeong-Sik Park , Ji-Hwan Kim and Gil-Jin Jang * SK Holdings C&C, Gyeonggi-do 13558, Korea; odh1501@gmail.com School of Electronic and Electrical Engineering, Kyungpook National University, Daegu 41566, Korea Department of English Linguistics and Language Technology, Hankuk University of Foreign Studies, Seoul 02450, Korea; parkjs@hufs.ac.kr Department of Computer Science and Engineering, Sogang University, Seoul 04107, Korea; kimjihwan@sogang.ac.kr School of Electronics Engineering, Kyungpook National University, Daegu 41566, Korea * Correspondence: gjang@knu.ac.kr; Tel.: +82-53-950-5517 Featured Application: Automatic speech recognition; chatbot; voice-assisted control; multimodal man-machine interaction systems. Abstract: Speech recognition consists of converting input sound into a sequence of phonemes, then finding text for the input using language models. Therefore, phoneme classification performance is a critical factor for the successful implementation of a speech recognition system. However, correctly distinguishing phonemes with similar characteristics is still a challenging problem even for state-of- the-art classification methods, and the classification errors are hard to be recovered in the subsequent language processing steps. This paper proposes a hierarchical phoneme clustering method to exploit more suitable recognition models to different phonemes. The phonemes of the TIMIT database are carefully analyzed using a confusion matrix from a baseline speech recognition model. Using automatic phoneme clustering results, a set of phoneme classification models optimized for the generated phoneme groups is constructed and integrated into a hierarchical phoneme classification method. According to the results of a number of phoneme classification experiments, the proposed Citation: Oh, D.; Park, J.-S.; hierarchical phoneme group models improved performance over the baseline by 3%, 2.1%, 6.0%, and Kim, J.-H.; Jang, G.-J. Hierarchical Phoneme Classification for Improved 2.2% for fricative, affricate, stop, and nasal sounds, respectively. The average accuracy was 69.5% Speech Recognition. Appl. Sci. 2021, and 71.7% for the baseline and proposed hierarchical models, showing a 2.2% overall improvement. 11, 428. https://doi.org/10.3390/ app11010428 Keywords: speech recognition; phoneme classification; clustering; recurrent neural networks Received: 2 December 2020 Accepted: 30 December 2020 Published: 4 January 2021 1. Introduction These days, automatic speech recognition (ASR) performance has improved greatly by Publisher’s Note: MDPI stays neu- using deep neural networks [1–5]. Since speech is a time series, recurrent neural networks tral with regard to jurisdictional clai- (RNNs) are mostly adopted for speech recognition [6–8], but there have been several ms in published maps and institutio- attempts to apply a convolution neural network (CNN) as well [9,10]. In most ASR systems, nal affiliations. speech is first converted to a sequence of phonemes, which is then translated to a word sequence using lexical models and language models generate meaningful sentences from the word sequence. Therefore, in ASR, the correct prediction of phonemes is the most Copyright: © 2020 by the authors. Li- critical factor because the errors in the earlier stage cannot be recovered in the final text. censee MDPI, Basel, Switzerland. Phonetically, the phonemes are distinguished according to the variations in pronunci- This article is an open access article ation and vocalization. However, most speech recognition methods do not actively consider distributed under the terms and con- different characteristics of phonemes, such as end-to-end speech recognition methods [11–16]. ditions of the Creative Commons At- Phonemes with similar characteristics are often confused and hard to distinguish, and tribution (CC BY) license (https:// they degrade the overall phoneme classification performance. Choosing feature extrac- creativecommons.org/licenses/by/ tion methods and acoustic model types appropriately for confusing phonemes can help 4.0/). Appl. Sci. 2021, 11, 428. https://doi.org/10.3390/app11010428 https://www.mdpi.com/journal/applsci Appl. Sci. 2021, 11, 428 2 of 17 improve the final sentence recognition performance. We propose a novel method of apply- ing phoneme-specific acoustic models for automatic speech recognition by a hierarchical phoneme classification framework. The hierarchical phoneme classification is composed of a single, baseline phoneme classifier, clustering into similar groups and final result generation using retrained group- specific models. First, we analyze the result of a general phoneme classifier based on long short-term memory (LSTM) [17]. In the next stage, we proceed clustering the confusing phonemes using a confusion matrix to group phonemes with similar characteristics. Then the phoneme groups are analyzed to find the characteristics that are common inside and different across the groups. Consonants and vowels are distinguished well and the length of consonants is usually shorter. In the proposed method, various lengths for the analysis frame are compared and the best one is chosen according to the phoneme recognition accuracy. In addition, there are several pairs of different consonants that differ by being voiced or unvoiced. For example, ‘d’ and ‘t’ sounds in different words ‘dean’ and ‘teen’, respectively, or ‘b’ and ‘p’ sounds in ‘bad’ and ‘pad’. Those consonants can be distinguished by the existence of the glottal pulse that occurs at periodic time intervals [18–20], and we use autocorrelation functions to add the periodicity feature of the phoneme sound if the found phoneme falls into consonant categories. The proposed hierarchical framework uses group-specific models with different features for different phoneme groups, trained by separate training data for each phoneme group. The rest of the paper is organized as follows: Section 2 describes the phoneme cluster- ing process from the confusion matrix of the baseline phoneme classification model and shows the resultant phoneme groups by the clustering. Section 3 explains a hierarchical phoneme classification model that adds a phoneme group model to the baseline phoneme classification and evaluate its performance. In Section 4, the evaluation results of the proposed method on phoneme recognition for the TIMIT dataset are given. 2. Phoneme Clustering This section describes the basic definition of phonemes, particularly in English, the phoneme classes used in this paper, and the baseline phoneme classification method using RNN. Then the pairwise misclassification rates between all the pairs of the phonemes are obtained to build a confusion matrix, and phoneme clustering results are obtained by grouping phonemes so that the phonemes within the group should be more confused than the phonemes between the groups. The phoneme clustering results are used in building the proposed hierarchical classifier in Section 3. 2.1. Phonemes Phonemes are the set of units representing the distinctive sounds of a spoken language. In most languages, there are about 20 to 60 units. In this paper, we adopt ARPAbet [21], developed by the Advanced Research Projects Agency (ARPA), which exploits standard English alphabet symbols to represent IPA (International Phonetic Alphabet) symbols [22]. Table 1 shows the mapping from ARPAbet to IPA symbols, with 11 more symbols [23] that are used in the TIMIT dataset [24,25] for a more detailed phoneme analysis. The symbol [h#] is not an actual phoneme but a special indicator symbol for the beginning and ending of a sentence, so the number of additional TIMIT phonemes is 11. The phonemes are broadly classified into consonants and vowels [26], and they are shown in the first and second columns of Table 1. The number of vowels and consonants used in ARPAbet are 19 and 31, so the total number of phonemes in this paper is 61. The consonants and vowels are further divided into more detailed categories according to the method of vocalization [27,28]. The most common types of consonants are stops, fricatives, affricates, nasals, and semivowels [29]. Stops are generated by the constriction and release of air in the mouth (closing-compression-release) [30], so there exists a certain duration called “stop gap” when the air is compressed inside the mouth. The stop sounds usually have large energies in low frequencies, roughly below 200 Hz. Fricatives have a Appl. Sci. 2021, 11, 428 3 of 17 looser constriction in the mouth, which allows producing some amount of friction at the point of contact. Most of their acoustic energy occurs at higher frequencies. Fricatives are divided into sibilants versus non-sibilants [31]. Sibilants tend to be louder than non- sibilants, and turbulence noise is stronger in sibilants than non-sibilants. Nasals have the same constriction as plosives except that air is allowed to pass through the nose but not through the mouth [32]. Nasals show the abrupt loss of overall energy because the nose is less efficient than the mouth in radiating the energy to the outside. Nasals involve a voiced feature in the frequency domain and in addition, they have formant frequencies like vowels. Table 1. English phonetic symbols of the extended 2-letter ARPAbet used in the TIMIT dataset with mapping to IPA (International Phonetic Alphabet) symbols. The number of vowels is 19 and the number of consonants is 31. The begin/end marker is ignored in the phoneme recognition. There are 11 more special symbols used in the TIMIT dataset, so the total number of phonemes is 61. Vowels (19) Consonants (31) TIMIT Extension (11 + 1) ARPAbet IPA ARPAbet IPA ARPAbet IPA Symbol Description Devoiced [@] aa A b b q P ax-h ([@]) ae æ ch Ù r ô eng Syllabic [N] ah 2 d d s s hv Voiced [h] ao O dh D sh S bcl [b] closure aw aU dx R t t dcl [d] closure ax @ el l th T gcl [g] closure axr Ä em m v v kcl [k] closure ay aI en n w w pcl [p] closure eh E f f wh û tcl [t] closure er Ç g g y j pau Pause Epenthetic ey eI hh h z z epi silence Begin/end ih I jh à zh Z h# marker ix 1 k k iy i l l ow oU m m oy OI n n uh U ng N uw u nx  R ux 0 p p The set of 61 phonemes are mapped to a smaller set of 39 labels [33] are shown in Table 2. In the new phoneme classes with multiple phonemes (classes 2, 5, 6, 8, 14, 15, 19, 20, 21, 28, 38, and 39), similar phonemes are grouped together because they are known to be hard to distinguish with acoustic features. All the closure sounds, pause, and glottal plosive ([q], equivalent to IPA [P]) are all grouped together to mapped phoneme 39. They are relatively weak, are often accompanies by other sounds, and are very hard to distinguish by only sound signals without surrounding phonemes. 2.2. Baseline Phoneme Recognition with TIMIT Dataset TIMIT corpus is designed to provide speech data for studies of acoustic-phonetic and speech recognition systems [24,25]. Phonetic transcripts for all sentences are provided in the TIMIT corpus distribution. A set of 61 phoneme labels from ARPAbet symbols are used in the transcripts. We subsequently collapsed the set of 61 labels into a smaller set of 39 labels in Table 2 during the scoring of performance of phoneme recognition. The baseline monophone classifier is developed by a recurrent neural network (RNN) [34–36]. From the input speech signals, mel-frequency cepstral coefficients (MFCCs) [37,38] are extracted from the input speech, their first and second order derivatives are computed, Appl. Sci. 2021, 11, 428 4 of 17 and the resultant 39 dimensional vector is obtained at every 10 milliseconds analyzed by 25-millisecond-overlapping windows. In this feature extraction process, all phonemes are pre-processed identically. The input to the baseline RNN is the 39-dimensional MFCC feature vectors and the number of output nodes is 39 one-hot vectors according to the 39 labels of the phonemes. Table 2. Mapping 61 phonemes to 39 phonemes for phoneme recognition. The grouping is based on the linguistic similarities between phonemes [33]. No ARPAbet No ARPAbet No ARPAbet No ARPAbet 1 iy 11 oy 21 ng eng 31 b 2 ih ix 12 aw 22 v 32 p 3 eh 13 ow 23 f 33 d 4 ae 14 er axr 24 dh 34 dx ax ah 5 15 l el 25 th 35 t ax-h 6 uw ux 16 r 26 z 36 g 7 uh 17 w 27 s 37 k 8 ao aa 18 y 28 zh sh 38 hh hv bcl pcl 9 ey 19 m em 29 jh 39 dcl tcl gcl kcl q 10 ay 20 n en nx 30 ch epi pau 2.3. Confusion Matrix A confusion matrix is a table of comparison results for the predicted and actual values to analyze intra-class and inter-class performances. It is used to evaluate the performance of supervised learning, such as classification in machine learning [39]. After training the baseline phoneme classification model, on a test set we evaluate phoneme classification accuracies and create a confusion matrix. Figure 1 shows the obtained confusion matrix from baseline phoneme classification results for 39 mapped phonemes. The y-axis is the real labels of the given phonemes, and the x-axis is the predicted labels from the baseline RNN. Each element of the matrix, denoted by C(i, j) where i and j are row and column numbers, respectively, is the rate of predicting phoneme j when the actual one is i. According to the definition of the confusion matrix, the diagonal elements (i = j) are the correct classification rates for individual phonemes and all off-diagonal elements (i 6= j) are ratios of incorrect classification pairs. For example, the first column of the the third row, C(1, 3), is brighter than the nearby components with the of the exception diagonal ones. It means that the phoneme [aw] is misclassified a lot to the phoneme [ae]. Among off-diagonal elements, ([uh], [ax]), ([uh], [ih]), ([n], [m]), and ([z], [s]) are examples of phoneme pairs with high misclassification rates. 2.4. Phoneme Clustering Using Confusion Matrix We clustered phonemes using the similarities defined by the sum of the off-diagonal elements in the confusion matrix as shown in Figure 1. We use an agglomerative method based on bottom up clustering, which starts from a unique cluster and ties up a pair one at a time [40]. To define the distance between a phoneme pair, we use the confusion error rates, obtained by one of the off-diagonal elements of the confusion matrix. For a newly created phoneme cluster, the new error rate is computed by the average of the merged phonemes. The process stops until the desired number of clusters remain. The results of hierarchical clustering are usually presented in a dendrogram [41]. As shown in Figure 2, the dendrogram is a tree diagram representing clusters obtained from hierarchical clustering. The y-axis shows phoneme labels and the phonemes are hierarchically paired with the height of the pair connection being the misclassification rate between the phonemes Appl. Sci. 2021, 11, 428 5 of 17 in the corresponding pair. The misclassification rate range from 0.0 to 1.0 along the x-axis. This method is called hierarchical clustering because paired phonemes are considered as a single phoneme after the pair is merged and all phonemes finally converge to a single cluster. The goal of the the proposed method is to improve the performance of phoneme recognition. Therefore, for the confusing phonemes, other classification method that is different from the baseline phoneme classifier is applied to further distinguish the phonemes in the same cluster. ���������������� �� �� �� �� �� ��� �� �� �� �� �� �� �� ��� �� �� �� �� ����� ��� �� �� �� ��� �� �� �� �� ��� � � � � � � � � � � � � � � �� �� �� �� � � �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� �� ���������� Figure 1. Confusion matrix from the results of the baseline phoneme classification. Each row of the matrix represents the instances in a truth and each column represents the instances in a predicted class. th dh hh dx cl ng jh ch zh uh ax ih eh uw aw ae er iy ey ow ay ao oy 0.0 0.2 0.4 0.6 0.8 1.0 Figure 2. Dendrogram of clustering. The x-axis shows a similarity of the clustered nodes. The higher the x-value, the more similar they are to each other. The y-axis is the phoneme label. This dendrogram shows that the consonants and vowels are well distinguished. Appl. Sci. 2021, 11, 428 6 of 17 In Figure 2, if we backtrack the clustering path from the finally merged cluster at the right end, all the phonemes are broadly divided into two clusters. The largest cluster in the upper side is a set of phonemes from [th] to [zh]. The lower largest cluster consists of phonemes from [uh] to [oy], most of which are vowel sounds. However, [l] and [r] are consonants and more specifically, belong to voiced alveolar sounds. The phonemes [w] and [y] are semivowels, belonging to voiced labio-velar and voiced palatal sounds, respectively. These sounds are between vowels and consonants, so we split the lower cluster into “Vowels” and “Mixed”, as shown in the first and second rows of Table 3. Among vowels, the phonetic classification is somewhat grouped according to the pronunciation position similarly to the conventional phonetic classification, but is not accurately distinguished. This is because the pronunciation of vowels vary according to the adjacent phonemes. All of the phonemes in the upper largest cluster are consonants. The individual consonants have different characteristics, so we divided the consonant set into four major groups. The lower four rows of Table 3 are the phoneme groups based on the dendrogram in Figure 2 using a 0.6 cut-off with some additional adjustments. The first group is a set of nasal sounds, [m], [n], and [ng]. The second one is a group of fricatives, sounds with air friction with longer pronunciation time. Voiced sounds ([z] [zh] [v] [dh]) and their corresponding unvoiced sounds ([s] [f] [th]) are grouped together. The voiceless glottal fricative sound, [hh], also belongs to this group. The third is the affricates group ([jh] [ch]), weak sounds with friction. The stop sounds such as [t], [d], [p], and [b] are closely clustered. The clustering result is similar to the basic phonetic categories. Our resultant clusters are analogous to the conventional linguistic categories, so we can confirm that it provides reasonable results. Table 3. Phoneme groups obtained by agglomerative hierarchical clustering on the TIMIT train- ing dataset. Phonemes of the Group Group (Written in ARPAbet) iy ih eh ae ax uw uh Vowels ao ey ay oy aw ow er Mixed l r w y Nasals m n ng Fricatives s z zh f v th dh hh Consonants Affricates jh ch Stops b d g k p t dx To confirm that the phoneme categories are well clustered, we drew a confusion matrix with the new phoneme groups in Table 3, and it is shown in Figure 3a. The phonemes are reordered so that all the phonemes in the same group are located adjacently. The red-colored rectangles indicate phoneme groups, and Figure 3b is the table below the figure is a confusion matrix with the phoneme groups as classification targets. The diagonal ones represent that the accuracy of the group labeling is very high, which implies that the different groups are well distinguished. In other words, the phonemes within the same group are similar and confusing, so we focus on improving the recognition performance of these phonemes inside the groups. Appl. Sci. 2021, 11, 428 7 of 17 (a) Phoneme grouping in the confusion matrix. vowel 0.94 0.02 0.02 0.01 0.00 0.01 mixed 0.02 0.92 0.01 0.01 0.00 0.04 Real nasal 0.07 0.09 0.78 0.01 0.00 0.04 fric 0.01 0.04 0.01 0.88 0.01 0.06 affr 0.00 0.01 0.00 0.07 0.80 0.12 stop 0.05 0.05 0.01 0.04 0.01 0.85 vowel mixed nasal fric affr stop Prediction (b) Confusion matrix of the phoneme groups. Figure 3. (a) The confusion matrix rearranged according to the phoneme clustering result. Each row of the matrix represents the instances in a truth while each column represents the instances in a predicted class. (b) Confusion matrix of phoneme groups obtained by phoneme clustering. 3. Hierarchical Phoneme Classification This section analyzes the characteristics of clustered phoneme groups. We explain the similarities and differences between phoneme groups, and the detailed implementation of group-specific models. By combining the group-specific phoneme classification models, a hierarchical phoneme classification framework is proposed. All individual group models are implemented by deep neural networks. The neural network automatically finds and recognizes the characteristics of speech signals. For sequential inputs such as speech signals, modeling temporal patterns is necessary. A recurrent neural network (RNN) is a type of neural network that can effectively learn temporal data. Therefore, phoneme classification model in this paper is built using RNN. More specifically, bi-directional long short-term memory (BLSTM) is adopted [8,42,43]. 3.1. Overall Architecture We propose a hierarchical phoneme classification model to improve the performance of phoneme classification. Figure 4 shows the overall model architecture. First, phoneme is predicted through the baseline phoneme classification model in which a whole phoneme is learned by using the same feature extraction method without distinguished characteristics. The baseline model was constructed using long short-term memory (LSTM) [17]. LSTM is Real Appl. Sci. 2021, 11, 428 8 of 17 an RNN cell that can avoid the problem of vanishing gradient using three gates of input, forget, and output. A typical LSTM uses a unidirectional structure that learns only by check- ing information from previous contexts. In the speech recognition, a bi-directional structure that learns both past and future information shows good performance [42–44]. Therefore, we implemented the model with the bi-directional LSTM (BLSTM) as a baseline system for phoneme recognition [45]. Three BLSTM layers were stacked and a fully-connected layer is added. At the output, the phoneme is classified using the softmax layer. Vowels vowel model Mixed: [l] [r] [w] [y] mixed model nasals feature extraction model by group Predicted baseline Phoneme feature different window Phoneme phoneme clas- wave extraction size fricatives sification Label (MFCC_gws) model affricates Consonant model autocorrelation stops model Figure 4. Overall model structure of the proposed hierarchical classification. Using the baseline phoneme classifier, phoneme-dependent features are extracted and different phoneme group models are applied to obtain refined phoneme classification results. The learning target of the predicted phoneme for the baseline model is the index of the phoneme group in Table 3. Then, the speech feature is re-extracted using the optimal window size for each phoneme group. In addition, autocorrelation is used together to dis- tinguish voiced and unvoiced phonemes. The extracted speech features are used as inputs to each phoneme group model. The phoneme group model finally predicts the phoneme. 3.2. Vowels and Mixed Phoneme Classification Two models for phoneme groups, ‘Vowel’ and ‘Mixed’ are trained by using group- specific training data. The ‘Vowel’ has 14 different labels as listed in Table 3, which is reduced from 39 phoneme labels of the baseline classifier. Out of the TIMIT training dataset, audio segments labeled by iy, ih, . . ., er are extracted and used in training ‘Vowel model’. To re-train ‘Mixed model’, training subsets labeled by l, r, w, and er are used. The model architecture is shown in Figure 5. Predicted Phoneme BLSTM FC MFCC Phonemes wave layers layer Figure 5. Phoneme classification model architecture for the baseline, as well as Vowel and Mixed groups. 3.3. Varying Analysis Window Sizes for Consonants Consonants are short and instantaneous, and vowels are longer and more consis- tent [46]. It would be better to shorten the analysis window size in feature extraction to more accurately distinguish short consonants. Therefore, we empirically decided appro- priate analysis window lengths differently for individual phoneme groups. The baseline phoneme recognition model extracts features using MFCC with the same conditions such Appl. Sci. 2021, 11, 428 9 of 17 as same window and shift sizes for all phoneme. In this process, a 16 kHz sampling rate and 25 ms hamming window is used. The window shift was set to 10 ms. However, conso- nants are very short and strong compared to vowels. Since the properties of phonemes are different, we compared the performance of phoneme recognition by extracting features while changing window and shift sizes for each group. The window sizes are 5 ms, 10 ms, 15 ms, 20 ms, 25 ms, and 30 ms. ‘MFCC_gws’ in Figure 4 is the MFCC features with varying window sizes. 3.4. Voiced and Unvoiced Consonants Classification Several pairs of different consonants are distinguished by being voiced or unvoiced. For voiced sounds, the glottis is closed or nearly closed, whereas for voiceless sounds it is wide open, with the distance between the folds actually amounting to only a fraction of a millimeter [18]. The voiced sounds carry periodic pulses generated by the glottis vibration, while almost no periodicity but turbulent noise is observed in the unvoiced sounds. Phonemes [b], [p], [d], and [t], all belong to the stops group in the phonetic classification. However [b], [d] are voiced phonemes and [p], [t] are unvoiced ones. Phonemes [z] and [s] are both fricatives, but [z] is voiced and [s] unvoiced. These pairs of examples are usually hard to distinguish because they are generated at the almost same place in the oral cavity area. Similarly, [p] and [b] are produced at the same place in the mouth and the only difference is being voiced or unvoiced. If we add a feature that helps distinguish voiced and unvoiced phonemes especially when recognizing consonants and create a model that can classify them, the performance may be improved. Therefore, we use autocorrelation functions to add the periodicity feature of the phoneme sound if the found phoneme falls into consonant categories Voiced and unvoiced sounds can be distinguished by pitch, or equivalently funda- mental periodicity values [47]. Autocorrelation measures the similarity function of the time lag with itself in the signal [19,20]. Due to the vibration of the vocal cords, the voiced sound is periodic and this vibration interval is pitch. Autocorrelation is a unique case of cross-correlation [48,49]. The cross-correlation measures the similarity of the two discrete signals, x[n] and y[n], defined by [50]: x[n + t]y[n] , (1) n=¥ where t is an integer time lag that takes values from f0,1,2, . . . ,g. If two discrete signals are identical, this process is defined as autocorrelation, which is given by: ac(x, t) = x[n + l]x[n] . (2) n=¥ Algorithm 1 is the pseudo code for the autocorrelation computation. The array x is the speech signal, N is the size of signal, and the autocorrelation function is denoted by ac. Figure 6a shows the waveform of a voiced consonant phoneme [d] and its autocorrelation computed by Algorithm 1. Likewise, Figure 6b is for an unvoiced consonant phoneme [t]. The autocorrelation of phoneme [d] shows nonzero values, while that of phoneme [t] has almost zero, meaning that there is no significant pitch. From these results, autocorrelation can help in distinguishing voiced and unvoiced phonemes. Appl. Sci. 2021, 11, 428 10 of 17 Algorithm 1 Autocorrelation H] ac array of autocorrelation N length of signal x for i 0 to N do sum 0 for j 0 to N 1 do sum sum + (x[j + 1] x[j]) end for ac[i] sum/N end for 3.5. Consonant Group Model Architectures As described in the previous section, we focus on making a better decision on con- sonants rather than on vowels. Consonants show various types of energy distributions in phoneme groups. Therefore, we implement appropriate models for individual conso- nant phoneme groups and verify their performance. MFCC_gws is a feature extraction method with group different window sizes, as well as autocorrelation features added to help distinguish voiced phonemes and unvoiced phonemes within the consonant group. By combining the two features [51], three phoneme group models are created. Figure 7 shows three different model configurations. Figure 7a simply concatenates MFCC_gws and the autocorrelation vector at the input unit and passes them to a BLSTM layer. A fully connected (FC) layer generates softmax outputs as many as the number phoneme to generate predictions for phoneme labels. Figure 7b has individual BLSTM layers for MFCC_gws and the autocorrelation vector, and concatenates the outputs to pass them through an FC layer. The advantage of output concatenation over input concatenation is that the mismatches of the different kinds of features, such as scale differences can be automatically adapted to each other. The concatenation at the BLSTM output layers is expected to be more reliable than Figure 7a. Figure 7c adds another BLSTM layer to the concatenated output of Figure 7b. These three configurations are used to construct the hierarchical phoneme classifier in Figure 4. 2 − 00 2 − 00 0 50 100 150 200 250 300 0 50 100 150 200 250 300 350 times times 1 − 000 1 − 000 2 − 000 0 50 100 150 200 250 300 0 50 100 150 200 250 300 350 times times (a) Voiced phoneme [d]. (b) Unvoiced phoneme [t]. Figure 6. Waveforms and autocorrelation functions of a voiced sound (phoneme [d]) and a unvoiced sound (phoneme [t]). Upper floats are waveforms and lower floats are computed autocorrelation functions. Relative amplitude amplitude Relative amplitude amplitude Appl. Sci. 2021, 11, 428 11 of 17 autocorrelation Predicted Phoneme BLSTM FC concat Phoneme wave layers layer MFCC_gws (a) Input concatenation. BLSTM autocorrelation outputs layers Predicted Phoneme FC concat Phoneme wave layer BLSTM MFCC_gws outputs layers (b) Concatenated bi-directional long short-term memory (BLSTM) and fully connected network (FCN). BLSTM autocorrelation outputs layers Predicted Phoneme BLSTM FC concat Phoneme wave layers layer BLSTM MFCC_gws outputs layers (c) Concatenated BLSTM, and BLSTM and FCN. Figure 7. Various types of the consonant group models with additional autocorrelation features. (a) input features are concatenated. (b) BLSTM outputs are cocatenated. (c) another BLSTM layer is added for the concatenated BLSTM outputs. 4. Experimental Results This section shows the analysis results of the phoneme group characteristics described in Section 2 and also presents comparison results of the performance of the phoneme group model with that of the baseline model. 4.1. TIMIT Database We have carried out speech recognition experiments on the TIMIT corpus [25], which is designed to provide speech data for studies of acoustic-phonetic and speech recognition systems. The speech data was recorded at Texas Instruments (TI), transcribed at the Massachusetts Institute of Technology (MIT), and prepared for CD-ROM production by the National Institute of Standards and Technology (NIST). The corpus includes the speech waveforms with phonetic transcriptions. TIMIT corpus has 6300 sentences, spoken by Appl. Sci. 2021, 11, 428 12 of 17 630 speakers, who spoke exactly 10 sentences each. There are three types of sentences. Dialect sentences (SA) were recorded from speakers with dialectal diversity. Phonetically- compact (SX) and phonetically-diverse sentences (SI) were created to model different situations of phoneme distributions. Phonetic transcripts for all sentences are provided in the TIMIT corpus distribution. The full phoneme transcript of an example sentence is shown in Table 4. All the numbers are in samples, so the length of the audio file in the example is 46,797/16,000 kHz = 2.924 s. Table 4. An example of the phonetic transcripts in the TIMIT corpus. The audio file is “dr1/rcjf0/sa1.wav” in the provided CD-ROM, and its sentence transcript is “sa1.txt” in the same directory. The phonetic transcript file is “sa1.phn”, and full timings of phonemes are provided as well. File Begin End Text/Phoneme She had your dark suit in greasy sa.txt 0 46797 wash water all year. 0 3050 h# 3050 4559 sh 4559 5723 ix sa1.phn 5723 6642 hv 6642 8772 eh 8772 9190 dcl 9190 10,337 jh . . . . . . . . . 40,313 42,059 y 42,059 43,479 ih 43,479 44,586 axr 44,586 46,720 h# The whole corpus is divided into non-overlapping training and testing sets. We use core test set [25] in an evaluating performance of phoneme recognition. The core test data has a total of 24 speakers, more specifically two male and one female from each of the eight dialect regions. The number of phoneme tokens are 177,080 in the training set and 9098 in the testing set. The number of tokens for each of the five phoneme groups is shown in Table 5. Table 5. Phoneme token counts of the TIMIT train and core test sets, grouped by phoneme classes. Phoneme Fricatives Affricates Stops Nasals Vowels/Mixed Groups Train set 21,424 2031 22,281 14,157 75,257 Core test set 1114 95 1114 736 3955 4.2. MFCC vs. Log Filter Bank Energy We compared the performance of two commonly-used features in speech recognition. The first one is mel-frequency cepstral coefficients (MFCCs) and the other one is log filterbank energies. Figure 8 shows the performance differences of four phoneme groups according to the change of window sizes. Blue lines are MFCC results and orange lines are the results of log filter bank energy features [37,38,52]. In most cases, blue lines are below orange lines, showing less phoneme recognition errors. Based on these results, we chose MFCC in the phoneme group models and the overall model. Appl. Sci. 2021, 11, 428 13 of 17 Fricatives 0.35 MFCC logfbank 0.30 0.25 0.20 0.15 5 10 15 20 25 30 Affricates MFCC 0.27 logfbank 0.26 0.25 0.24 0.23 5 10 15 20 25 30 Window size [ms] (a) Fricatives and Affricates. Stops 0.40 MFCC logfbank 0.35 0.30 0.25 0.20 5 10 15 20 25 30 Nasals MFCC 0.30 logfbank 0.28 0.26 5 10 15 20 25 30 Window size [ms] (b) Stops and Nasals. Figure 8. The error rates using mel-frequency cepstral coefficients (MFCCs) and log filter bank energy according to change window size. (a) shows fricatives and afficates. (b) shows stops and nasals. In all phoneme groups, the error rate of MFCC is lower than that of log filter bank energy. 4.3. Various Window Sizes In order to find the optimal window sizes for individual phoneme groups, we tried various window sizes in the MFCC feature extraction. Table 6 shows performance varia- tions by choosing different window sizes with six phoneme group models. For fricatives and stops, 10 ms window size was the best, both 15 ms and 25 ms were the best for af- fricates ([jh] and [ch]) and 20 ms was the best for nasal. This is because fricative and stop sounds are very short consonants, and nasal sounds are voiced and close to vowel sounds. Generally, consonant models showed lower phoneme error rates with smaller window sizes because the actual vocalization length is shorter for consonants, so it is possible to obtain more information by using a small window length. Surprisingly, for vowel/semivowel and closes, window analysis size did not affect performance. Unlike consonants, they have very stable characteristics over time, so a change in analysis sizes error rate error rate error rate error rate Appl. Sci. 2021, 11, 428 14 of 17 made no difference. We used the best window sizes for each of the phoneme group models to maximize the performance of the proposed hierarchical phoneme classification model. 4.4. Phoneme Group Model Training The baseline model uses MFCC features only. Clustering result showed that voiced and unvoiced phonemes are hard to distinguish. To overcome this problem, autocorrelation feature is added in phoneme group models because it shows very different values for voiced and unvoiced speech signals. Table 7 shows the performance of phoneme group models. This result shows that using autocorrelation is better than the MFCC alone. In addition, the performance of the three models was different for each phoneme group. Since the phoneme group has different characteristics, the optimal model structure is also different. Based on these results, we constructed a hierarchical model using phoneme group models. 4.5. Performance of the Hierarchical Classification We implemented a hierarchical phoneme classification model that combines all the phoneme group models with the optimal window sizes for the corresponding phoneme groups and evaluated the performance. We used the Adam optimizer, which adaptively adjusts the learning rate of the backpropagation learning algorithm for the artificial neural networks [53]. The initial learning rate of the Adam optimizer is set to be 10 , and gradually lowered as the learning proceeds. Batch size 512 is used during training of each phoneme group model. Dropout [54] with a probability of 0.2 is added across the BLSTM layers. Table 8 represents the performance of the overall model. Compared with the baseline, models with different window sizes for phoneme groups had an improved performance. In addition, the overall model using the phoneme group model combining MFCC and autocorrelation showed better performance. Table 6. The results of overall model using MFCC with varying window sizes. Analysis Window Sizes Baseline 5 ms 10 ms 15 ms 20 ms 25 ms 30 ms Vowels 0.700 - - - - - - Mixed 0.919 - - - - - - Nasals 0.568 0.565 0.553 0.564 0.579 0.557 0.568 Fricatives 0.724 0.710 0.739 0.738 0.737 0.735 0.736 Affcicates 0.632 0.632 0.642 0.653 0.590 0.653 0.642 Stops 0.628 0.657 0.682 0.651 0.647 0.626 0.620 Group 0.695 0.697 0.706 0.704 0.695 0.698 0.698 avg Integrated 0.730 0.732 0.738 0.735 0.735 0.731 0.731 Table 7. The results that different window sizes in MFCC and MFCC with autocorrelation and a three feature ensemble model. MFCC/Autocorrelation Combinations Phoneme Group Specific Concat BLSTM Concat BLSTM Input Concat + BLSTM + Groups Window Sizes + FCN FCN Nasals 0.742 0.757 0.742 0.757 Fricatives 0.844 0.847 0.834 0.827 Affricates 0.768 0.737 0.779 0.747 Stops 0.796 0.781 0.806 0.778 Appl. Sci. 2021, 11, 428 15 of 17 Table 8. The results that baseline and integration phoneme group model. Overall Model Phoneme Baseline Group MFCC and Groups Phoneme Specific Autocorrealation Classification Window Size Combination Model Vowels 0.700 0.700 0.700 Mixed 0.919 0.919 0.919 Nasals 0.568 0.579 0.590 Fricatives 0.724 0.739 0.754 Affcicates 0.632 0.653 0.653 Stops 0.628 0.682 0.688 Group avg 0.695 0.712 0.717 Integrated 0.730 0.740 0.743 5. Discussion In this paper, we proposed a hierarchical speech recognition model based on phoneme clustering. The proposed model first classifies the entire phoneme by a baseline phoneme classifier without considering phoneme characteristics. Then the retrained phoneme group models are used using the initial phoneme classification results. A confusion matrix of baseline phoneme classification was used to group each phoneme with similar charac- teristics. The confusion matrix shows which phonemes are not well distinguished. An analysis of phoneme characteristics in the TIMIT database revealed the differences of the phonemes in terms of lengths and autocorrelation values, so they were used in constructing phoneme group-specific models. We extracted features using different window size and window shift optimized for each phoneme group, and used autocorrelation as a feature as well as MFCC to improve the performance of the phoneme group models. The results of the experiment of the four consonant phoneme group models showed that the phoneme groups have different characteristics, and thus, the suitable models are different. The hier- archical phoneme classification model consisting of the baseline model and the phoneme group model was evaluated. The result shows that the phoneme specific feature extraction method and classification model have an improved performance of phoneme classification and speech recognition. One drawback of the proposed hierarchical classification method is that it does not improve the vowel classification performance because the proposed autocorrelation features and varying window sizes only help consonant classification. The MFCC is well suited to vowel classification and almost optimal classification performance is achieved, so very little improvement is expected. Author Contributions: Conceptualization, J.-S.P., J.-H.K., and G.-J.J.; methodology, D.O.; software, D.O.; validation, D.O. and G.-J.J.; formal analysis, J.-H.K. and J.-S.P.; investigation, J.-H.K.; resources, G.-J.J. and J.-H.K.; data curation, D.O.; writing–original draft preparation, D.O. and G.-J.J.; writing– review and editing, J.-S.P. and J.-H.K.; visualization, D.O.; supervision, G.-J.J.; project administration, G.-J.J.; funding acquisition, G.-J.J. All authors have read and agreed to the published version of the manuscript. Funding: This work was supported by Institute for Information and Communications Technology Promotion (IITP) grant funded by the Korean government (MSIT) (No. R7124-16-0004, Development of Intelligent Interaction Technology Based on Context Awareness and Human Intention Understand- ing), by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. NRF-2017M3C1B6071400), and the BK21 Plus project funded by the Ministry of Educa- tion, Korea (21A20131600011). Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Appl. Sci. 2021, 11, 428 16 of 17 Data Availability Statement: The TIMIT dataset can be obtained from the Linguistic Data Consor- tium (LDC), https://www.ldc.upenn.edu. Conflicts of Interest: The authors declare no conflict of interest. Abbreviations The following abbreviations are used in this manuscript: ARPA Advanced Research Projects Agency ASR Automatic Speech Recognition BLSTM Bi-directional Long Short-Term Memory CNN Convolutional Neural Network IPA International Phonetic Alphabet LSTM Long Short-Term Memory MFCC Mel-Frequency Cepstral Coefficient NIST National Institute of Standards and Technology RNN Recurrent Neural Network References 1. Hinton, G.; Deng, L.; Yu, D.; Dahl, G.E.; Mohamed, A.R.; Jaitly, N.; Senior, A.; Vanhoucke, V.; Nguyen, P.; Sainath, T.N.; et al. Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups. IEEE Signal Process. Mag. 2012, 29, 82–97. [CrossRef] 2. Yu, D.; Li, J. Recent progresses in deep learning based acoustic models. IEEE/CAA J. Autom. Sin. 2017, 4, 396–409. 3. Xiong, W.; Wu, L.; Alleva, F.; Droppo, J.; Huang, X.; Stolcke, A. The Microsoft 2017 conversational speech recognition system. In Proceedings of the 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Calgary, AB, Canada, 15–20 April 2018; pp. 5934–5938. 4. Caterini, A.L.; Chang, D.E. Deep Neural Networks in a Mathematical Framework; Springer: New York, NY, USA, 2018. [CrossRef] 5. Taylor, S.; Kim, T.; Yue, Y.; Mahler, M.; Krahe, J.; Rodriguez, A.G.; Hodgins, J.K.; Matthews, I.A. A deep learning approach for generalized speech animation. ACM Trans. Gr. 2017, 36, 1–11. [CrossRef] 6. Graves, A. Sequence Transduction with Recurrent Neural Networks. arXiv 2012, arXiv:1211.3711. 7. Graves, A.; Mohamed, A.R.; Hinton, G. Speech recognition with deep recurrent neural networks. In Proceedings of the 2013 IEEE International Conference on Acoustics, Speech and Signal Processing, Vancouver, BC, Canada, 26–31 May 2013; pp. 6645–6649. 8. Zeyer, A.; Doetsch, P.; Voigtlaender, P.; Schlüter, R.; Ney, H. A Comprehensive Study of Deep Bidirectional LSTM RNNs for Acoustic Modeling in Speech Recognition. In Proceedings of the 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), New Orleans, LA, USA, 5–9 May 2017; pp. 2462–2466. 9. Abdel-Hamid, O.; Mohamed, A.R.; Jiang, H.; Deng, L.; Penn, G.; Yu, D. Convolutional neural networks for speech recognition. IEEE/ACM Trans. Audio Speech Lang. Process. 2014, 22, 1533–1545. [CrossRef] 10. Tóth, L. Phone recognition with hierarchical convolutional deep maxout networks. EURASIP J. Audio Speech Music Process. 2015, 2015, 25. [CrossRef] 11. Hannun, A.; Case, C.; Casper, J.; Catanzaro, B.; Diamos, G.; Elsen, E.; Prenger, R.; Satheesh, S.; Sengupta, S.; Coates, A.; et al. Deep speech: Scaling up end-to-end speech recognition. arXiv 2014, arXiv:1412.5567. 12. Miao, Y.; Gowayyed, M.; Metze, F. EESEN: End-to-end speech recognition using deep RNN models and WFST-based decoding. In Proceedings of the 2015 IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU), Scottsdale, AZ, USA, 13–17 December 2015; pp. 167–174. 13. Bahdanau, D.; Chorowski, J.; Serdyuk, D.; Brakel, P.; Bengio, Y. End-to-end attention-based large vocabulary speech recognition. In Proceedings of the 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Shanghai, China, 20–25 March 2016; pp. 4945–4949. 14. Zhang, Y.; Pezeshki, M.; Brakel, P.; Zhang, S.; Bengio, C.L.Y.; Courville, A. Towards end-to-end speech recognition with deep convolutional neural networks. arXiv 2017, arXiv:1701.02720. 15. Battenberg, E.; Chen, J.; Child, R.; Coates, A.; Gaur, Y.; Li, Y.; Liu, H.; Satheesh, S.; Seetapun, D.; Sriram, A.; et al. Exploring Neural Transducers for End-to-End Speech Recognition. CoRR 2017, abs/1707.07413. 16. Chiu, C.C.; Sainath, T.N.; Wu, Y.; Prabhavalkar, R.; Nguyen, P.; Chen, Z.; Kannan, A.; Weiss, R.J.; Rao, K.; Gonina, E.; et al. State-of-the-art speech recognition with sequence-to-sequence models. In Proceedings of the 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Calgary, AB, Canada, 15–20 April 2018; pp. 4774–4778. 17. Hochreiter, S.; Schmidhuber, J. Long short-term memory. Neural Comput. 1997, 9, 1735–1780. [PubMed] 18. Hayes, B. Introductory Phonology; Malden, M.A., Ed.; Wiley-Blackwell: Oxford, UK, 2009. 19. Rabiner, L. On the use of autocorrelation analysis for pitch detection. IEEE Trans. Acoust. Speech Signal Process. 1977, 25, 24–33. [CrossRef] Appl. Sci. 2021, 11, 428 17 of 17 20. Hernández, M. A tutorial to extract the pitch in speech signals using autocorrelation. Open J. Technol. Eng. Discip. (OJTED) 2016, 2, 1–11. 21. Klautau, A. ARPABET and the TIMIT Alphabet. 2001. 22. International Youth Association TIP. Reproduction of The International Phonetic Alphabet; Cambridge University Press: Lon- don, UK, 2005. 23. Gold, B.; Morgan, N.; Ellis, D. Speech and Audio Signal Processing: Processing and Perception of Speech and Music, 2nd ed.; Wiley: Hoboken, NJ, USA, 2011. 24. Linguistic Data Consortium. Table of All the Phonemic and Phonetic Symbols Used in the Timit Lexicon; Linguistic Data Consortium: Philadelphia, PA, USA, 1990. 25. Garofolo, J.; Lamel, L.; Fisher, W.; Fiscus, J.; Pallett, D. DARPA TIMIT acoustic-phonetic continous speech corpus CD-ROM. NIST speech disc 1-1.1. NASA STI/Recon Tech. Rep. N 1993, 93, 1–94. 26. Ladefoged, P.; Disner, S.F. Vowels and Consonants; John Wiley & Sons: Hoboken, NJ, USA, 2012. 27. Ades, A.E. Vowels, consonants, speech, and nonspeech. Psychol. Rev. 1977, 84, 524. [CrossRef] 28. Stevens, K.N. Acoustic Phonetics; MIT Press: Cambridge, MA, USA, 2000; Volume 30, 29. Miller, G.A.; Nicely, P.E. An analysis of perceptual confusions among some English consonants. J. Acoust. Soc. Am. 1955, 27, 338–352. [CrossRef] 30. Halle, M.; Hughes, G.W.; Radley, J.P. Acoustic properties of stop consonants. J. Acoust. Soc. Am. 1957, 29, 107–116. [CrossRef] 31. Shadle, C.H. The acoustics of fricative consonants. Tech. Rep. 1985, 1–200. [CrossRef] 32. Fujimura, O. Analysis of nasal consonants. J. Acoust. Soc. Am. 1962, 34, 1865–1875. [CrossRef] 33. Halberstadt, A.K. Heterogeneous Acoustic Measurements and Multiple Classifiers for Speech Recognition; Massachusetts Institute of Technology: Cambridge, MA, USA, 1999. 34. Evangelopoulos, G.N. Efficient Hardware Mapping of Long Short-Term Memory Neural Networks for Automatic Speech Recognition. Master ’s Thesis, KU Leuven, Belgium, 2016. 35. Ek, J.V.; Michálek, J.; Psutka, J. Recurrent DNNs and its Ensembles on the TIMIT Phone Recognition Task. arXiv 2018, arXiv:1806.07186. 36. Michaálek, J.; Ek, J.V. A Survey of Recent DNN Architectures on the TIMIT Phone Recognition Task. arXiv 2018, arXiv:1806.07974. 37. Han, W.; Chan, C.F.; Choy, C.S.; Pun, K.P. An efficient MFCC extraction method in speech recognition. In Proceedings of the 2006 IEEE International Symposium on Circuits and Systems, Island of Kos, Greece, 21–24 May 2006; p. 4. 38. Muda, L.; Begam, M.; Elamvazuthi, I. Voice recognition algorithms using mel frequency cepstral coefficient (MFCC) and dynamic time warping (DTW) techniques. arXiv 2010, arXiv:1003.4083. 39. Davis, J.; Goadrich, M. The relationship between Precision-Recall and ROC curves. In Proceedings of the 23rd International Conference on Machine Learning, Pittsburgh, PA, USA, 25–29 June 2006; pp. 233–240. 40. Corpet, F. Multiple sequence alignment with hierarchical clustering. Nucl. Acids Res. 1988, 16, 10881–10890. [CrossRef] 41. Wilks, D.S. Cluster analysis. In International Geophysics; Elsevier: Amsterdam, The Netherlands, 2011; Volume 100, pp. 603–616. 42. Schuster, M.; Paliwal, K.K. Bidirectional recurrent neural networks. IEEE Trans. Signal Process. 1997, 45, 2673–2681. [CrossRef] 43. Graves, A.; Schmidhuber, J. Framewise phoneme classification with bidirectional LSTM networks. In Proceedings of the 2005 IEEE International Joint Conference on Neural Networks, Montreal, QC, Canada, 31 July–4 August 2005; Volume 4, pp. 2047–2052. 44. Hannun, A.Y.; Maas, A.L.; Jurafsky, D.; Ng, A.Y. First-pass large vocabulary continuous speech recognition using bi-directional recurrent DNNs. arXiv 2014, arXiv:1408.2873. 45. Graves, A.; Fernández, S.; Schmidhuber, J. Bidirectional LSTM networks for improved phoneme classification and recognition. In International Conference on Artificial Neural Networks; Springer: New York, NY, USA, 2005; pp. 799–804. 46. Caramazza, A.; Chialant, D.; Capasso, R.; Miceli, G. Separable processing of consonants and vowels. Nature 2000, 403, 428. [CrossRef] [PubMed] 47. Rabiner, L.R.; Schafer, R.W. Digital Processing of Speech Signals; Pearson Publishing: NewYork, NY, USA, 1978; Volume 100. 48. Chen, C.H. Signal Processing Handbook; CRC Press: Boca Raton, FL, USA, 1988; Volume 51. 49. Furui, S.; Sondhi, M.M. Advances in Speech Signal Processing; Marcel Dekker, Inc.: New York, NY, 1991. 50. Proakis, J.G. Digital Signal Processing: Principles Algorithms and Applications; Pearson Education India Services Pvt.: Chennai, India, 51. Ngiam, J.; Khosla, A.; Kim, M.; Nam, J.; Lee, H.; Ng, A.Y. Multimodal deep learning. In Proceedings of the 28th International Conference On Machine Learning (ICML-11), Bellevue, WA, USA, 28 June–2 July 2011; pp. 689–696. 52. Nadeu, C.; Hernando, J.; Gorricho, M. On the decorrelation of filter-bank energies in speech recognition. In Proceedings of the Fourth European Conference on Speech Communication and Technology, Madrid, Spain, 18–21 September 1995. 53. Kingma, D.P.; Ba, J. Adam: A method for stochastic optimization. arXiv 2014, arXiv:1412.6980. 54. Srivastava, N.; Hinton, G.; Krizhevsky, A.; Sutskever, I.; Salakhutdinov, R. Dropout: A simple way to prevent neural networks from overfitting. J. Mach. Learn. Res. 2014, 15, 1929–1958.

Journal

Applied SciencesMultidisciplinary Digital Publishing Institute

Published: Jan 4, 2021

There are no references for this article.