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

Learn More →

Performance of Adaptive Noise Cancellation with Normalized Last-Mean-Square Based on the Signal-to-Noise Ratio of Lung and Heart Sound Separation

Performance of Adaptive Noise Cancellation with Normalized Last-Mean-Square Based on the... Hindawi Journal of Healthcare Engineering Volume 2018, Article ID 9732762, 10 pages https://doi.org/10.1155/2018/9732762 Research Article Performance of Adaptive Noise Cancellation with Normalized Last-Mean-Square Based on the Signal-to-Noise Ratio of Lung and Heart Sound Separation Noman Q. Al-Naggar and Mohammed H. Al-Udyni Department of Biomedical Engineering, Faculty of Engineering, University of Science and Technology, Sana’a, Yemen Correspondence should be addressed to Noman Q. Al-Naggar; noman_qaed@yahoo.com Received 16 January 2018; Revised 3 May 2018; Accepted 31 May 2018; Published 12 July 2018 Academic Editor: Olivier Beuf Copyright © 2018 Noman Q. Al-Naggar and Mohammed H. Al-Udyni. /is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. /e adaptive algorithm satisfies the present needs on technology for diagnosis biosignals as lung sound signals (LSSs) and accurate techniques for the separation of heart sound signals (HSSs) and other background noise from LSS. /is study investigates an improved adaptive noise cancellation (ANC) based on normalized last-mean-square (NLMS) algorithm. /e parameters of ANC-NLMS al- gorithm are the filter length (L ) parameter, which is determined in 2 sequence of 2, 4, 8, 16, . . . , 2048, and the step size (μ ), which is j n automatically randomly identified using variable μ (VSS) optimization. Initially, the algorithm is subjected experimentally to identify the optimal μ range that works with 11 L values as a specific case. /is case is used to study the improved performance of the n j proposed method based on the signal-to-noise ratio and mean square error. Moreover, the performance is evaluated four times for four μ values, each of which with all L to obtain the output SNR matrix (4 × 11). /e improvement level is estimated and compared with n j out the SNR prior to the application of the proposed algorithm and after SNR . /e proposed method achieves high-performance in outs ANC-NLMS algorithm by optimizing VSS when it is close to zero at determining L , at which the algorithm shows the capability to separate HSS from LSS. Furthermore, the SNR of normal LSS starts to improve at L of 64 and L limit of 1024. /e SNR of out j j out abnormal LSS starts from a L value of 512 to more than 2048 for all determined μ . Results revealed that the SNR of the abnormal j n out LSS is small (negative value), whereas that in the normal LSS is large (reaches a positive value). Finally, the designed ANC-NLMS algorithm can separate HSS from LSS. /is algorithm can also achieve a good performance by optimizing VSS at the determined 11 L values. Additionally, the steps of the proposed method and the obtained SNR may be used to classify LSS by using a computer. out other interference signals or noises. /e adaptive noise can- 1. Introduction celler (ANC) used in this study is a type of AF. Lung sound signals (LSSs) exhibit nonperiodicity and low Many works have widely investigated the filtering and frequency; these signals also contain symptoms of many separation of LSS by using the ANC or the adaptive line diseases and interfere with frequency components (50– enhancement (ALE) with the last-mean-square (LMS) and 2500 Hz) with heart sound signal (HSS) frequency in the range normalized last-mean-square (NLMS) algorithms [2–7]. of 20–600 Hz [1]. Furthermore, the interference between LSS NLMS algorithm can be used to separate HSS from LSS [3, 4] and HSS is high due to the nearby positions and physiological because it can deal with two signals recorded in real time. In recording points of the two signal sources. /erefore, the general, previous studies have focused on the main pa- keeping symptoms on LSS overlap and the increase in diffi- rameters of AF, including the filter length (L), constant step culty of separating HSS and other noise from LSS. /ey require size (μ ), filter type (such as ALE or ANC), and algorithm modern and highly accurate tools for filtering and separation. (such as NLMS or LMS) to obtain improved AF perfor- /e adaptive filter (AF) satisfies the LSS purification re- mance. However, these parameters and combination of quirements, and it is an effective tool used to filter LSS from techniques have been used with several limitations. 2 Journal of Healthcare Engineering /e effect and estimating performance of the designed 0.8 method were studied using power spectrum density (PSD), 0.6 which is based on monitoring the concentration of an av- 0.4 erage power frequency. /e PSD graphic shows the com- 0.2 parison before and after signal separation [4–6]. A few studies have also investigated the effects of separating dif- ferent biosignals from noises, such as LSS, ECG, and –0.2 myoelectric signals, on the signal-to-noise ratio (SNR) at –0.4 specified requirement outputs [3, 8–11]. –0.6 /e present study evaluates the estimation performance of –0.8 0 2000 4000 6000 8000 10000 12000 14000 16000 ANC based on NLMS algorithm to separate HSS from con- Sample number (n) taminated LSS on the SNR and the behavior of mean square error (MSE). Moreover, the improvement in performance level Figure 1: Heart sound signal. is studied under four values of the optimal variable μ (VSS) and 11 determined L values in the following 2 sequence: j � 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048. /erefore, the performance for one separation is processed 44 times with NLMS to separate HSS from LSS. /e NLMS is more (4μ × 11), that is, the SNR is calculated to obtain 4 × 11 matrix stable than the LMS in terms of dealing with more than one of the output SNR values. Such combination of the proposed out signal in real-time applications; the NLMS algorithm also algorithm overcomes the limitations of previous studies in displays higher and faster rate of convergence than that of addition to the following: the use of NLMS algorithm instead of LMS [12]. According to the stated abilities, NLMS algorithm LMS algorithm because LMS algorithm cannot be adopted with is used in this case study. /e original input of ANC is used two long signals and the use of ANC instead of ALE. /e VSS for contaminated LSS, and the reference input is used for the initially is studied to identify the optimal range that can work noise HSS. Figure 2 illustrates the main components of with 11 L . /e level of performance improvement is estimated ANC-NLMS algorithm architecture. by comparing the SNR before and after applying the proposed /e inputs of ANC-NLMS represent two wave files, each method. /e proposed method is carried out and processed of which is recorded by an individual channel. /e original using a code program on the MATLAB platform. /e proposed signal X (n) is contaminated by the reference signal h (n) i i method can deal with large data, process repeatedly according during the recording process. to the number of the L values, and calculate the SNR values. j out /e original input signal X (n) can be described as Results revealed the ability of the designed ANC-NLMS follows: algorithm to separate HSS from LSS successfully and showed the increasing performance with increasing L value. /e X (n) � d (n) + h (n), (1) j i i i0 improved SNR of the normal and abnormal LSSs is par- where d (n) is the desired pure lung sound (LSS), h (n) is ticularly achieved at the L range of 64–1024 and 512–2048, i i0 the interfered HSS in X (n) that represents noise, and i is respectively, at the determined μ . /e comparison of SNR n in a corresponding order number of the signal. with the obtained matrix of the SNR aids in exploring the out existence of distinguishable characteristics between normal (i) /e reference input signal, that is, h (n), is assumed and abnormal LSSs, which can be used in computerized LSS almost correlated with h (n). i0 classification. /e filter output Y (n) is defined as follows: L−1 2. Materials and Methods Y (n) � 􏽘 w (n) ∗ X(n − k) i k 2.1. Materials. Required signals of heart and lung sounds k�0 (2) used for experiments are obtained from the laboratory of the � w (n)x(n) (estimate of d(n)), Biomedical Engineering Department at University of Sci- ence and Technology, Yemen. LSS and HSS are recorded where L is the filter length, and j is the value determined using two-channel electronic stethoscopes and stored in wav from the 2 sequence of 2, 4, 8, 16, 32, 64, 128, 256, 512,1024, format [4]. Both signals are recorded with sampling fre- 2048, at which the designed algorithm performance is ex- quency of 44100 Hz. HSS recording is carried out on the amined. Additionally, k is a number of iteration, x(n) � [x(n)· down anterior region of the chest. /e HSS used in this study x(n − 1) . . . x(n − L − 1)] is the input vector of time delayed is normal, which consists of the first heart sound (S1) and input values, and w(n) � [w (n) · w (n − 1) . . . w (n)] 0 1 L−1 murmurs (shown in Figure 1). is the weight vector at the time n that can be minimized, as /e lung sound auscultation is performed on the left shown in down posterior and right anterior regions of the chest. /e 2 2 (3) lung sounds considered in this study are described in Table 1. ‖w(n)‖ � ‖w(n + 1) · · · w(n)‖ . 2.2. ANC Algorithm Architecture. /e present study in- (ii) /e μ value for the input vector is calculated as vestigates on the performance of the combination of ANC follows: Amplitude Journal of Healthcare Engineering 3 Table 1: Lung sound data. N Name Type of sound Status Recording position SNR (db) in 1 LSN1 Vesicular Normal Posterior: left, low −8.65 2 LSA2 Crackles Abnormal Posterior: left, middle −14.4 3 LSN3 Bronchial Normal Chest: right, up −3.93 4 LSA4 Wheeze Abnormal Posterior: left, middle −15.9 5 LSN5 Broncho-vesicular Normal Posterior: left up −7.78 6 LSA6 Crackles Abnormal Posterior: right, low −53.8 Table 2: Summary of the NLMS algorithm. Tap-weight vector, w(n), Input: Input vector, x(n), and desired output, d(n) Output: Filter output, y(n), tap-weight vector update, w(n + 1) 1. Filtering output signal: y(n) � w (n)x(n) 2. Error estimation: e(n) � d(n) − y(n) 3. Tap weight and step size parameters adaptation: w(n + 1) � w(n) + α(x(n)/β + x(n) ) × e(n) α Original signal, X (n)= d(n)+ h (n) e(n) Output i 0 Lung μ � � � , � �2 (4) � � sound β + �X � n – where β is a small positive constant used to avoid division by zero when the input vector X is zero. /us, the problem on y(n) Heart Reference signal obtaining a gradient noise amplification in tap weights is ANC–NLMS sounds h (n) solved. Furthermore, α is the adaptation positive constant that is commonly less than 1 (0 < α < 1) [12, 13]. Figure 2: ANC-NLMS algorithm architecture. 2.3. NLMS Optimization. NLMS optimization is a principal approaches, that is, two parameters (α and β) will be method for minimal disturbance presented in [13], where controlled to satisfy the required performance. /e the error signal e (n) is defined as the difference between the i experiments are carried out with consideration of the desired signal and the filter output in (2). Hence, the error is following: minimized in magnitude and rearranged as follows: (a) /e adaptation constant α is changed within the e (n) � d (n) − Y (n) � d (n) − w (n + 1)x(n). (5) i i i i range of 0 < α < 1, and the small positive constant β is changed within the range 0.1–0.009 [12]. /e NLMS algorithm recursion obtains the constrained (b) /e VSS is studied within the range of 0-1 at the optimization criterion. /e tap weight is as follows: determined L value. x(n) (c) /e influence on the overall performance is w(n + 1) � w(n) + α × e(n). (6) β + x(n) monitored on the minimization of MSE, SNR out behavior, and algorithm output graphics. NLMS algorithm is an indication of the minimal dis- (ii) Second step: auto-optimum VSS turbance among iterations [13, 14]. Table 2 summarizes the NLMS algorithm. /e proposed idea here is a modified method from pseudorandom number generator μ for NLMS al- gorithm [16]. /e main parameters α and β are 2.4. NLMS μ . /e μ parameter should be optimized to n n changed randomly into variable value from random ensure the reliability of the designed algorithm [15] at 11 L numbers of distribution from 0 to 1 at each iteration values (determined previously) as a case study. /e optimal time. μ is obtained within a fixed optimal range of μ is obtained through the following steps. First, most ideal 0 ≥ μ � 0.1, which is explored experimentally from VSS is randomly searched. Results from the first step are the first step. /e proposed idea is implemented, as used in the second step. Such results include the imple- shown in Table 3; it achieves the optimum solution of mentation and automation of the algorithm work. Both steps NLMS in Section 2.3. are described in further detail in the following paragraphs. (i) First step: random search for the most ideal 2.5. Performance Analysis possible μ μ presents two main parameters, namely, α and β, 2.5.1. MSE. MSE is a performance function of AF, and its which can affect the overall performance of the al- target is the low MSE value for it to achieve a proper per- gorithm. /is aspect is the motivation for the VSS formance [13]. /erefore, the values and graph of this 4 Journal of Healthcare Engineering Table 3: Summary of optimum ANC-NLMS algorithm for HSS Table 4: Calculated MSE during searching for the optimal VSS. cancellation. MSE For time index, n � 1, 2,. . ., L filter length L (number of iteration) # µ L � 4 16 64 128 256 1024 and j � [2, 4, 8,. . ., 2048] 1 0.6 0.00003 Inf NaN NaN NaN NaN /e number of L value (j), 2 0.2 0.00210 0.00679 NaN NaN NaN NaN N (1, . . . , 4) the number of step size 3 0.1 0.00329 0.00329 Inf NaN NaN NaN Tap-weight vector, w(n), 4 0.09 0.00336 0.00590 2052.75 Inf NaN NaN Input Input vector, x(n) 5 0.041 0.00347 0.00500 0.00490 0.00864 Inf NaN Desired output, d(n) 6 0.011 0.00291 0.00155 0.00865 0.00423 0.00475 0.01501 Alpha � rand(1, N) 7 0.009 0.00278 0.00028 0.00150 0.00371 0.00413 0.01327 Beta � rand(1, N) Filter output, y(n) Output Tap-weight vector update, w(n + 1) 1. Filtering y(n) � w (n)x(n) E􏼂Y (n)􏼃 2. Error estimation e(n) � d(n) − y(n) SNR (dB) � 10 log 10 , (9) out µ ·L􏼁 n j For i � 0 : L − 1 E e (n) 􏼂 􏼃 For j � 0: N − 1 m(j) � mu/(x(n)^2 + be) where Y (n) is the output (pure LSS) of ANC-NLMS and If m(j) > mu max considered the signal, and e (n) is the estimated error (noise 3. Step size m(j) � mu max measurement) of ANC-NLMS and considered the noise. /e calculation If m(j) < mu min higher output SNR (SNR ) than that of SNR indicates the out in m(j) � mu min pureness of the obtained LSS and success of the noise removal End and consequently improves the performance of ANC-NLMS. End /e improvement level is estimated as follows: 4. Tap weight and step size 2 SNR (dB) � SNR − SNR . (10) imro. out in w(n + 1) � w(n) + μ(x(n)/β + x(n) ) × e(n) parameters adaptation 2.5.3. Output Graphics. Visual graphics are used as metrics in observing the change in input and output graphics. /ese quantity are essential to evaluate the performance of the AF. graphics will illustrate the input signals (original and ref- /e formula for MSE is given by the following equation: erence) in two windows and two other windows for output MSE(n) � E􏽮e (n)􏽯, (7) signals (pure LSS and estimated error). Accordingly, the change can be easily observed. where E[·] denotes the statistical expectation, and e is the /e experiment is carried out using MATLAB platform, estimated error of AF. /e MSE is calculated for the evo- in which an algorithm code is designed to obtain the main lution of AF performance during searching for the optimal output signals, their graphic matrix (SNR ) (11) SNR , and out in VSS, as shown in Table 4. other input parameters. SNR SNR SNR · · · SNR 􏼂 􏼃 􏼂 􏼃 􏼂 􏼃 􏼂 􏼃 out out out out 2.5.2. SNR Evaluation. SNR is used as a metric to estimate ⎢ ⎥ ⎡ ⎢ ⎤ ⎥ ⎢ ⎥ ⎢ μ (1),ℓ �2 μ (1),ℓ �4 μ (1),ℓ �8 μ (1),ℓ �2048 ⎥ ⎢ n j n j n j n j ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ the performance of the proposed method, and it is defined as ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 􏼂SNR 􏼃 􏼂SNR 􏼃 􏼂SNR 􏼃 · · · 􏼂SNR 􏼃 ⎥ ⎢ ⎥ ⎢ out out out out ⎥ the ratio of the amount of signal to the amount of noise [17]. ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ μ (2),ℓ �2 μ (2),ℓ �4 μ (2),ℓj�8 μ (2),ℓ �2048 ⎥ ⎢ ⎥ ⎢ n j n j n n j ⎥ ⎢ ⎥ ⎢ ⎥ In the present study, SNR is calculated before and after ⎢ ⎥ ⎢ ⎥ ⎢ ⎥, (11) ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ applying the ANC-NLMS algorithm to compare their values ⎢ 􏼂SNR 􏼃 􏼂SNR 􏼃 􏼂SNR 􏼃 · · · 􏼂SNR 􏼃 ⎥ ⎢ ⎥ ⎢ out out out out ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ μ (3),ℓ �2 μ (3),ℓ �4 μ (3),ℓ �8 μ (3),ℓ �2048 ⎥ ⎢ ⎥ ⎢ n j n j n j n j ⎥ for the same signals at the determining condition. /e input ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ SNR (SNR ) of the recording signal is measured in am- ⎢ ⎥ ⎢ ⎥ in ⎣ ⎦ SNR SNR SNR · · · SNR 􏼂 􏼃 􏼂 􏼃 􏼂 􏼃 􏼂 􏼃 out out out out plitudes; thus, SNR must be squared to be proportional to μ (4),ℓ �2 μ (4),ℓ �8 μ (4),ℓ �8 μ (4),ℓ �2048 in n j n j n j n j power, as expressed in (8) [18, 19]. SNR is used as one of metrics for the improvement of AF out E􏼂X (n)􏼃 performance during searching for the optimal VSS. /e (8) SNR (dB) � 10 log 10 , in E􏼂h (n)􏼃 obtained SNR is shown in Table 5. where X (n) is the original signal defined in (1) and con- sidered the signal, h (n) is the reference signal and considered 2.6. Experiment Procedures. /e experiment procedures are the noise, and i refers to the same number of pair signals. summarized as follows: Moreover, E(·) denotes operations in calculating the ex- (i) Create the coding program. pectation calculation in the time domain. According to the (ii) Unite the frequency sampling (8000 Hz). proposed method, (8) is suitable for SNR calculation because h (n) is correlated with existing noise (h (n)) in X (n). (iii) /e maximum duration of studied signal is 3.5 s, i 0 i /e output SNR value after applying ANC-NLMS is that is, one completed breathing cycle, which is out given by the following equation: equal to 28000 samples. Journal of Healthcare Engineering 5 Table 5: A sample of searching the optimal µ for the designed algorithm (SNR � −7.78). n in SNR (dB) out # µ L � 4 8 16 32 64 128 256 512 1024 2048 1 0.8 −5.3 −3.4 −1.8 −0.4 2.1 0.0 NaN NaN NaN NaN 2 0.6 −6.1 −4.1 −2.4 −0.9 1.4 0.0 NaN NaN NaN NaN 3 0.35 −8.0 −5.7 −3.8 −2.2 −0.1 3.0 0.0 NaN NaN NaN 4 0.26 −9.1 −6.6 −4.6 −2.9 −0.9 2.3 6.3 NaN NaN NaN 5 0.178 −10.7 −8.0 −5.7 −3.9 −2.0 1.1 5.5 7.5 NaN NaN 6 0.09 −13.5 −10.7 −8.1 −5.8 −3.9 −0.9 3.6 6.4 7.6 NaN 7 0.0797 −13.9 −11.1 −8.4 −6.2 −4.2 −1.2 3.3 6.2 7.5 0.0 8 0.0088 −24.5 −20.9 −17.6 −14.6 −11.8 −8.8 −5.0 −1.4 1.6 3.8 Learning curve for MSE –16 –17 –18 –19 –20 –21 –22 –23 –24 –25 –26 0 200 400 600 800 1000 1200 Number of iterations (n) mu = 0.0285 mu = 0.085 mu = 0.0334 mu = 0.063 mu = 0.0100 Figure 3: MSE performance of the NLMS for various optimal μ values and when L � 64. (iv) Experimentally identify the optimal μ range as were considered because of their effects on the performance stated in Section 2.5. of the designed algorithm. Figure 3 shows the MSE of μ with a value of 0.06, which (v) Procedure is performed with μ for each L value n 1 j displays faster convergence rate than those of others. Addi- (i.e., 11 times according to the j values) to calculate tionally, the AF became steady after approximately 200 it- and obtain the SNR 11 times at each L value. out j erations at steady state error of approximately −24 dB. /e (vi) /e procedure is repeated similarly with each μ other MSE tools needed a long time to converge and be- value, that is, four values within the determined came steady after approximately 400 iterations at steady optimal μ value, to obtain 44 total processing for state errors of approximately −26 dB for μ � 0.033 and signal at (μ · L ), where j � 2, 4, 8, 16, . . . , 2048. n j −25.5 dB for μ � 0.028 and 0.0085. /us, the steady state /erefore, SNR is calculated 44 times and updated out errors were small. with each μ to obtain the matrix shown in (11). ni Figure 4 illustrates the results for large VSS that results in (vii) /e experiment is carried out on MATLAB platform, unstable performance and unsatisfied results. According to in which an algorithm code is designed to obtain the the comparison between Figures 3 and 4, the performance main output signals and performance analysis tools. was good when VSS was small and close to zero. /e same conclusion was observed in the changes in MSE and SNR out values; they improved gradually with decreasing μ and 3. Results when they became close to zero, as shown in Tables 4 and 5. To obtain reliable results during all procedures, including the Searching for the optimal VSS identified the VSS optimal searching for the optimal VSS, the number of samples and L range of 0 ≥ μ � 0.1. /us, the designed algorithm lost its j n Mean (error ) (dB) 6 Journal of Healthcare Engineering Learning curve for MSE 0.02 0.018 0.016 0.014 0.012 0.01 0.008 0.006 0.004 0 200 400 600 800 1000 1200 Number of iterations (n) mu = 0.1261 mu = 0.1161 Figure 4: MSE performance of NLMS for μ > 0.100 values and L � 64. stability when μ was used without the identified range, as started from L � 64 and increased with increased L value. n j j shown in Tables 4 and 5 and Figures 3 and 4. /e improvement level based on SNR also increased with out Figures 5(a)–5(c) display that the maximum amplitude increased L value and obtained small change at different of input signal (Figure 5(a)) is 0.18, the maximum amplitude VSS. of the error signal is approximately 0.0015 (Figure 5(c)), and /e SNR values of normal LSS ranged from −3.93 dB to in the maximum amplitude of square error is 0.00004 (Figure 5(b)). −7.78 dB (Table 1). /e SNR values changed according to out /is result showed an increase in the algorithm number of the L and determined μ (Table 7). j n computation windows and minimization of error, as well Figure 7 exhibits the signal obtained by the designed as MSE, which is considered a function of NLMS perfor- algorithm, with the visual difference between the original mance. /erefore, NLMS optimization achieved minimal signal (graphic 1, Figure 7(a)) and the pure LSS (graphic 3, disturbance, and the designed algorithm accurately adapted Figure 7(c)). In general, the frequency components were low and converged to separate HSS from the original signal in the subjected original signal and decreased after applying (Section 2.3). AF on the pure LSS (graphic 3, Figure 7(c)). /e auto-optimal algorithm of identifying VSS has been used to evaluate the separation of HSS from the original 4. Discussions signal that consists of contaminated LSS and HSS based on SNR, as well as the performance of ANC and NLMS al- /e optimized results of the designed algorithm determined gorithm combination. the optimal VSS range of 0 ≥ μ � 0.1; in this range, the AF Table 6 shows the SNR values calculated from the AF became highly stable with nonstationary biosignals, such out outputs for the abnormal LSS case. Results of the com- LSS, where the performance of the proposed method showed parison of the SNR values shown in Table 1 and the SNR the most ideal trade-off between convergence speed and low in out values shown in Table 6 indicated that the SNR values were steady error on the basis of the appropriately autoselected µ in located from −14.4 dB to −53 dB in an abnormal case. /is [7, 15]. /is achievement approved the proper work of the result suggested that the abnormal LSS included high designed algorithm and its capability to separate signals by amount of noise, and the SNR values changed according identifying the VSS range on NLMS algorithm, which is in out to the L and determined VSS. agreement with the results of several works [20–22]. Figure 6 demonstrates the visual difference before and /e SNR values improved progressively at determined out after applying the proposed method at the determined pa- μ and L values for abnormal and normal lung sounds, as n j rameters where the original signal (graphic 1, Figure 6(a)) shown in Tables 6 and 7, respectively. /e SNR matrix out showed higher frequency components than that of pure LSS indicated that the performance level in the normal and (graphic 3, Figure 6(c)). In addition to the improved level of abnormal LSSs started improving from the L values of SNR , these results indicated the separation of the noise 64,128 and 256,512, respectively. Moreover, L can be out j�1024 components from the desired LSS. the upper limit at which the AF may work stably with normal Table 7 shows the improved performance level of sep- LSS and obtain accurate outputs. AF can work at L more arating signals on normal LSS case by observing SNR that than 2048 with abnormal LSS. /is result can be due to that it out Mean (error ) (dB) Journal of Healthcare Engineering 7 0.2 0.1 –0.1 –0.2 –0.3 0 0.5 1 1.5 2 2.5 ×10 (a) –4 ×10 0 0.5 1 1.5 2 2.5 ×10 Sample number (n) (b) –3 ×10 1.5 0.5 0 0.5 1 1.5 2 2.5 ×10 (c) Figure 5: Comparison of the amplitude of AF original signal X (n) and obtained errors. (a) Original signal. (b) MSE of the NLMS in the sample when μ � 0.09 and L � 128. (c). Output error of the NLMS. Table 6: SNR after applying the designed algorithm for an abnormal LSS case. L values LSS# µ 2 4 8 16 32 64 128 256 512 1024 2048 SNR (dB) out 0.111 −45.43 −38.31 −34.66 −16.16 −24.08 −19.57 −15.43 −15.09 −11.04 −5.06 −1.05 0.036 −45.43 −40.01 −32.09 −29.44 −13.39 −20.63 −16.99 −15.34 −3.65 −7.40 −2.16 LSA2 0.022 −28.12 −40.01 −32.75 −29.44 −24.66 −18.93 −13.87 −10.41 −3.41 −2.84 −3.08 0.016 −45.43 −40.01 −19.56 −19.31 −24.66 −17.19 −16.99 −15.34 −11.04 −7.95 −3.95 0.037 −35.28 −42.26 −40.03 −34.64 −16.15 −24.17 −17.89 −15.35 −14.04 −11.12 −3.80 0.028 −51.46 −45.66 −37.56 −34.64 −26.35 −24.40 −11.74 −13.11 −13.46 −11.12 −4.45 LSA4 0.025 −51.46 −44.18 −39.50 −31.80 −29.32 −13.80 −19.17 −18.29 −14.04 −3.80 −4.79 0.010 −37.72 −41.70 −40.03 −34.64 −16.15 −23.30 −20.41 −18.29 −6.95 −7.40 −8.00 0.100 −106.11 −101.49 −95.90 −83.13 −82.24 −79.34 −75.47 −56.85 −42.00 −43.41 −32.31 0.039 −101.38 −113.39 −89.27 −83.13 −86.10 −83.84 −65.40 −48.74 −41.56 −36.56 −40.06 LSA6 0.037 −101.38 −115.35 −109.27 −103.13 −89.80 −87.29 −82.06 −58.56 −57.56 −47.07 −40.58 0.017 −102.41 −95.35 −95.59 −100.87 −80.84 −70.46 −64.75 −48.74 −57.63 −36.11 −46.87 consists of normal LSS for low-frequency components, and between both LSS types whether in terms of SNR or SNR in out the abnormal LSS consists of high-frequency components. values. /e SNR values revealed several distinct markers /e performance estimation of ANC-NLMS algorithm out between LSSs; the normal LSS shows high SNR values, that combination based on automatic identification of the op- is, the SNR value reaches close to 0 or the positive axis. By timal VSS validated the correctness of the proposed contrast, the SNR values of the abnormal LSS are consid- method and its sequence steps in separating HSS from erably small in the negative axis. /erefore, these charac- LSS. Additionally, such estimation explored the distinct teristics may reveal clear difference that can differentiate features differentiating normal LSS from abnormal LSS, Mean (error ) 8 Journal of Healthcare Engineering –3 e original LSS corrupted by HSS ×10 –2 (a) e heart sound signal as reference –1 (b) e lung sound aer adaptive filter (Y) 0.2 –0.2 (c) e error –1 (d) Figure 6: Input and output signal graphics of ANC-NLMS algorithm on an abnormal case (LSA6). Obtained graphics and SNR � −40.58 out at μ � 0.039 and L � 2048, where the input SNR is −53.86 dB. Table 7: SNR after applying the designed algorithm on normal LSS. L values LSS# µ 2 4 8 16 32 64 128 256 512 1024 2048 SNR (dB) out 0.064 −24.82 −16.36 −16.82 −13.63 −11.06 −8.52 −6.18 −2.24 0.83 3.76 17.64 0.021 −24.82 −20.59 −16.82 −13.63 −11.06 −6.51 −7.00 −2.24 2.86 3.58 12.39 LSN1 0.017 −24.82 −20.59 −16.82 −8.11 −10.66 −8.18 −3.34 −0.93 3.52 3.73 11.25 0.010 −24.75 −20.59 −16.82 −11.11 −10.85 −9.20 −7.00 −2.24 3.46 3.58 8.83 0.025 −20.90 −18.84 −16.76 −14.79 −13.37 −9.60 −9.49 −8.26 −3.23 −3.68 −0.63 0.011 −21.17 −16.92 −16.76 −13.31 −12.31 −12.28 −9.49 −2.63 −4.94 −3.68 −1.64 LSN3 0.010 −21.17 −17.52 −16.76 −14.79 −13.42 −12.46 −4.64 −5.35 −5.81 −3.68 −1.42 0.009 −21.17 −18.84 −16.68 −14.79 −13.37 −11.90 −8.21 −2.68 −5.81 −3.68 −1.95 0.026 −25.95 −24.07 −19.72 −17.32 −9.96 −10.56 −8.47 −4.76 −0.14 1.78 6.56 0.020 −24.56 −13.45 −19.11 −12.52 −12.21 −10.51 −8.47 −4.76 2.15 5.02 6.02 LSN5 0.013 −27.94 −21.41 −18.37 −17.32 −14.24 −10.29 −8.47 −4.76 −0.78 1.78 4.88 0.010 −26.35 −19.43 −18.55 −17.32 −9.71 −8.15 −8.47 −4.76 −1.10 1.78 3.94 and these characteristic may be used as primary features to VSS for 11 L values to separate HSS from LSS. /e per- classify LSS. formance of the designed algorithm evaluated at determined conditions showed good result by reducing and minimizing the error gradually to zero after the convergence time. 5. Conclusions /e effectiveness of the designed algorithm to separate /is study investigated an effective method of ANC-NLMS HSS from contaminated LSS estimated based on the SNR out algorithm based on automatic identification of the optimal illustrated a progressive performance improvement level Journal of Healthcare Engineering 9 e original LSS corrupted by HSS 0.5 –0.5 (a) e heart sound signal as reference –1 (b) e lung sound aer adaptive filter (Y) –1 (c) e error –1 (d) Figure 7: Input and output signal graphics of ANC-NLMS algorithm on a normal case (LSN1). /e obtained graphics and SNR at out μ � 0.017, where the SNR is −8.65 dB, and the SNR is −3.34 dB at L � 128. n in out [3] K. Sathesh and N. J. Muniraj, “Real time heart and lung sound with increasing L and significantly improved separation of separation using adaptive line enhancer with NLMS,” Journal HSS from LSS. of ;eoretical and Applied Information Technology, vol. 65, /is SNR explored a novel method to differentiate out no. 2, pp. 559–564, 2014. between normal and abnormal LSSs. /is method may be [4] N. Q. Al Naggar and H. Ghazi, “Design two-channel in- used as basis in developing computerized diagnosis and strument to record lung and heart sounds at the same time, automating LSS calcification. and separate them using ANC-NLMS algorithm,” In- /e proposed approach clarified the correctness of the ternational Journal of Advanced Research in Electrical, Elec- combined designed algorithm and achieved significant tronics and Instrumentation Engineering, vol. 4, no. 4, performance. /e proposed method may be subject for pp. 2601–2609, 2015. further study on LSS under different settings and durations. [5] T. Tsalaile and S. Sanei, “Separation of heart sound signal from lung sound signal by adaptive line enhancement,” in Pro- Data Availability ceedings of the 15th European Signal Processing Conference, vol. 15, pp. 1231–1235, Poznan, Poland, 2007. /e data used to support the findings of this study are [6] N. Q. Al-Naggar, “A new method of lung sounds filtering available from the corresponding author upon request. using modulated least mean square adaptive noise cancella- tion,” Journal of Biomedical Science and Engineering, vol. 6, no. 9, pp. 869–876, 2013. Conflicts of Interest [7] R. Nersisson and M. M. Noel, “Hybrid nelder-mead search based optimal least mean square algorithms for heart and lung /e authors declare that they have no conflicts of interest. sound separation,” Engineering Science and Technology, an International Journal, vol. 20, no. 3, pp. 1054–1065, 2017. References [8] S. Sebastian and S. Rathnakara, “Separation of heart sound artifact from respiratory signals using singular spectrum based [1] S. Reichert, R. Gass, C. Brandt et al., “Analysis of respiratory advanced line enhancer,” International Journal on Advanced sounds: state of the art,” Clinical Medicine Circulatory, Re- Computer ;eory and Engineering, vol. 2, pp. 106–111, 2013. spiratory and Pulmonary Medicine, vol. 2, pp. 45–58, 2008. [9] M. Molaie, S. Jafari, M. H. Moradi, J. C. Sprott, and [2] R. K. /enua and S. K. Agarwal, “Simulation and performance analysis of adaptive filter in noise cancellation,” International S. M. R. H. Golpayegani, “A chaotic viewpoint on noise re- duction from respiratory sounds,” Biomedical Signal Pro- Journal of Engineering Science and Technology, vol. 9, pp. 4373–4378, 2010. cessing and Control, vol. 10, pp. 245–249, 2014. 10 Journal of Healthcare Engineering [10] O. El B’charri, R. Latif, K. Elmansouri, A. Abenaou, and W. Jenkal, “ECG signal performance de-noising assessment based on threshold tuning of dual-tree wavelet transform,” Biomedical Engineering Online, vol. 16, no. 1, 2017. [11] H. T. Ma and Y. T. Zhang, “Effects of the physiological pa- rameters on the signal-to-noise ratio of single myoelectric channel,” Journal of Neuroengineering and Rehabilitation, vol. 4, no. 1, p. 29, 2007. [12] K. Sathesh, N. J. R. Muniraj, A. V. Akshitha, B. K. Roy, M. Induja, and M. D. A. Maharasi, “Design and imple- mentation of real time electronic stethoscope with a method for separating heart sound from lung sound using TMS320C6713,” DSK Middle-East Journal of Scientific Re- search, vol. 23, no. 8, pp. 1909–1920, 2015. [13] B. Farhang-Boroujeny, Adaptive Filters ;eory and Applications, University of Utah, Wiley & Sons, Somerset, NJ, USA, 2nd edition, 2013. [14] S. Zhao, “Performance analysis and enhancements of adaptive algorithms and their applications of computer engineering,” Doctor of Philosophy thesis, Nanyang Technological Uni- versity, Singapore, 2009. [15] S. S. Haykin, Adaptive Filter ;eory, Pearson Education, New York, Ny, USA, 4th edition, 2008. [16] S. Jimaa, “Convergence evaluation of a random step-size NLMS adaptive algorithm in system identification and channel equalization,” in Adaptive Filtering, InTech, London, UK, 2011. [17] J. L. Semmlow, Biosignal and Biomedical Image Processing: MATLAB-Based Applications, Marcel Dekker, New York, NY, USA, 2004. [18] B. Y. Lu, “Unidirectional microphone based wireless recorder for the respiration sound,” Journal of Bioengineering and Biomedical Science, vol. 6, no. 3, p. 195, 2016. [19] A. H. Said, “Design of a chopper amplifier for use in bio- medical signal acquisition,” Master thesis, Department of Electrical and Computer Engineering, Southern Illinois University, Carbondale, IL, USA, 2010. [20] R. M. A. Falcão, “Adaptive filtering algorithms for noise cancellation,” Masters thesis, Faculdade de Engenharia da Universidade do Porto, Porto, Portugal, 2012. [21] S. A. Jimaa, N. Al Saeedi, S. Al-Araji, and R. M. Shubair, “Performance evaluation of random step-size NLMS in adaptive channel equalization,” in Proceedings of the 2008 1st IFIP Wireless Days, Dubai, UAE, November 2008. [22] K. Gomathi, V. Saravanan, and N. Santhiyakumari, “Variable step size for improving convergence of FxLMS,” Procedia Technology, vol. 25, pp. 420–426, 2016. International Journal of Advances in Rotating Machinery Multimedia Journal of The Scientific Journal of Engineering World Journal Sensors Hindawi Hindawi Publishing Corporation Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 http://www www.hindawi.com .hindawi.com V Volume 2018 olume 2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Journal of Control Science and Engineering Advances in Civil Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Submit your manuscripts at www.hindawi.com Journal of Journal of Electrical and Computer Robotics Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 VLSI Design Advances in OptoElectronics International Journal of Modelling & Aerospace International Journal of Simulation Navigation and in Engineering Engineering Observation Hindawi Hindawi Hindawi Hindawi Volume 2018 Volume 2018 Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com www.hindawi.com www.hindawi.com Volume 2018 International Journal of Active and Passive International Journal of Antennas and Advances in Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Healthcare Engineering Hindawi Publishing Corporation

Performance of Adaptive Noise Cancellation with Normalized Last-Mean-Square Based on the Signal-to-Noise Ratio of Lung and Heart Sound Separation

Download PDF
11 pages

Loading next page...
 
/lp/hindawi-publishing-corporation/performance-of-adaptive-noise-cancellation-with-normalized-last-mean-0QG767MdaH

References (23)

Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2018 Noman Q. Al-Naggar and Mohammed H. Al-Udyni. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
ISSN
2040-2295
eISSN
2040-2309
DOI
10.1155/2018/9732762
Publisher site
See Article on Publisher Site

Abstract

Hindawi Journal of Healthcare Engineering Volume 2018, Article ID 9732762, 10 pages https://doi.org/10.1155/2018/9732762 Research Article Performance of Adaptive Noise Cancellation with Normalized Last-Mean-Square Based on the Signal-to-Noise Ratio of Lung and Heart Sound Separation Noman Q. Al-Naggar and Mohammed H. Al-Udyni Department of Biomedical Engineering, Faculty of Engineering, University of Science and Technology, Sana’a, Yemen Correspondence should be addressed to Noman Q. Al-Naggar; noman_qaed@yahoo.com Received 16 January 2018; Revised 3 May 2018; Accepted 31 May 2018; Published 12 July 2018 Academic Editor: Olivier Beuf Copyright © 2018 Noman Q. Al-Naggar and Mohammed H. Al-Udyni. /is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. /e adaptive algorithm satisfies the present needs on technology for diagnosis biosignals as lung sound signals (LSSs) and accurate techniques for the separation of heart sound signals (HSSs) and other background noise from LSS. /is study investigates an improved adaptive noise cancellation (ANC) based on normalized last-mean-square (NLMS) algorithm. /e parameters of ANC-NLMS al- gorithm are the filter length (L ) parameter, which is determined in 2 sequence of 2, 4, 8, 16, . . . , 2048, and the step size (μ ), which is j n automatically randomly identified using variable μ (VSS) optimization. Initially, the algorithm is subjected experimentally to identify the optimal μ range that works with 11 L values as a specific case. /is case is used to study the improved performance of the n j proposed method based on the signal-to-noise ratio and mean square error. Moreover, the performance is evaluated four times for four μ values, each of which with all L to obtain the output SNR matrix (4 × 11). /e improvement level is estimated and compared with n j out the SNR prior to the application of the proposed algorithm and after SNR . /e proposed method achieves high-performance in outs ANC-NLMS algorithm by optimizing VSS when it is close to zero at determining L , at which the algorithm shows the capability to separate HSS from LSS. Furthermore, the SNR of normal LSS starts to improve at L of 64 and L limit of 1024. /e SNR of out j j out abnormal LSS starts from a L value of 512 to more than 2048 for all determined μ . Results revealed that the SNR of the abnormal j n out LSS is small (negative value), whereas that in the normal LSS is large (reaches a positive value). Finally, the designed ANC-NLMS algorithm can separate HSS from LSS. /is algorithm can also achieve a good performance by optimizing VSS at the determined 11 L values. Additionally, the steps of the proposed method and the obtained SNR may be used to classify LSS by using a computer. out other interference signals or noises. /e adaptive noise can- 1. Introduction celler (ANC) used in this study is a type of AF. Lung sound signals (LSSs) exhibit nonperiodicity and low Many works have widely investigated the filtering and frequency; these signals also contain symptoms of many separation of LSS by using the ANC or the adaptive line diseases and interfere with frequency components (50– enhancement (ALE) with the last-mean-square (LMS) and 2500 Hz) with heart sound signal (HSS) frequency in the range normalized last-mean-square (NLMS) algorithms [2–7]. of 20–600 Hz [1]. Furthermore, the interference between LSS NLMS algorithm can be used to separate HSS from LSS [3, 4] and HSS is high due to the nearby positions and physiological because it can deal with two signals recorded in real time. In recording points of the two signal sources. /erefore, the general, previous studies have focused on the main pa- keeping symptoms on LSS overlap and the increase in diffi- rameters of AF, including the filter length (L), constant step culty of separating HSS and other noise from LSS. /ey require size (μ ), filter type (such as ALE or ANC), and algorithm modern and highly accurate tools for filtering and separation. (such as NLMS or LMS) to obtain improved AF perfor- /e adaptive filter (AF) satisfies the LSS purification re- mance. However, these parameters and combination of quirements, and it is an effective tool used to filter LSS from techniques have been used with several limitations. 2 Journal of Healthcare Engineering /e effect and estimating performance of the designed 0.8 method were studied using power spectrum density (PSD), 0.6 which is based on monitoring the concentration of an av- 0.4 erage power frequency. /e PSD graphic shows the com- 0.2 parison before and after signal separation [4–6]. A few studies have also investigated the effects of separating dif- ferent biosignals from noises, such as LSS, ECG, and –0.2 myoelectric signals, on the signal-to-noise ratio (SNR) at –0.4 specified requirement outputs [3, 8–11]. –0.6 /e present study evaluates the estimation performance of –0.8 0 2000 4000 6000 8000 10000 12000 14000 16000 ANC based on NLMS algorithm to separate HSS from con- Sample number (n) taminated LSS on the SNR and the behavior of mean square error (MSE). Moreover, the improvement in performance level Figure 1: Heart sound signal. is studied under four values of the optimal variable μ (VSS) and 11 determined L values in the following 2 sequence: j � 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048. /erefore, the performance for one separation is processed 44 times with NLMS to separate HSS from LSS. /e NLMS is more (4μ × 11), that is, the SNR is calculated to obtain 4 × 11 matrix stable than the LMS in terms of dealing with more than one of the output SNR values. Such combination of the proposed out signal in real-time applications; the NLMS algorithm also algorithm overcomes the limitations of previous studies in displays higher and faster rate of convergence than that of addition to the following: the use of NLMS algorithm instead of LMS [12]. According to the stated abilities, NLMS algorithm LMS algorithm because LMS algorithm cannot be adopted with is used in this case study. /e original input of ANC is used two long signals and the use of ANC instead of ALE. /e VSS for contaminated LSS, and the reference input is used for the initially is studied to identify the optimal range that can work noise HSS. Figure 2 illustrates the main components of with 11 L . /e level of performance improvement is estimated ANC-NLMS algorithm architecture. by comparing the SNR before and after applying the proposed /e inputs of ANC-NLMS represent two wave files, each method. /e proposed method is carried out and processed of which is recorded by an individual channel. /e original using a code program on the MATLAB platform. /e proposed signal X (n) is contaminated by the reference signal h (n) i i method can deal with large data, process repeatedly according during the recording process. to the number of the L values, and calculate the SNR values. j out /e original input signal X (n) can be described as Results revealed the ability of the designed ANC-NLMS follows: algorithm to separate HSS from LSS successfully and showed the increasing performance with increasing L value. /e X (n) � d (n) + h (n), (1) j i i i0 improved SNR of the normal and abnormal LSSs is par- where d (n) is the desired pure lung sound (LSS), h (n) is ticularly achieved at the L range of 64–1024 and 512–2048, i i0 the interfered HSS in X (n) that represents noise, and i is respectively, at the determined μ . /e comparison of SNR n in a corresponding order number of the signal. with the obtained matrix of the SNR aids in exploring the out existence of distinguishable characteristics between normal (i) /e reference input signal, that is, h (n), is assumed and abnormal LSSs, which can be used in computerized LSS almost correlated with h (n). i0 classification. /e filter output Y (n) is defined as follows: L−1 2. Materials and Methods Y (n) � 􏽘 w (n) ∗ X(n − k) i k 2.1. Materials. Required signals of heart and lung sounds k�0 (2) used for experiments are obtained from the laboratory of the � w (n)x(n) (estimate of d(n)), Biomedical Engineering Department at University of Sci- ence and Technology, Yemen. LSS and HSS are recorded where L is the filter length, and j is the value determined using two-channel electronic stethoscopes and stored in wav from the 2 sequence of 2, 4, 8, 16, 32, 64, 128, 256, 512,1024, format [4]. Both signals are recorded with sampling fre- 2048, at which the designed algorithm performance is ex- quency of 44100 Hz. HSS recording is carried out on the amined. Additionally, k is a number of iteration, x(n) � [x(n)· down anterior region of the chest. /e HSS used in this study x(n − 1) . . . x(n − L − 1)] is the input vector of time delayed is normal, which consists of the first heart sound (S1) and input values, and w(n) � [w (n) · w (n − 1) . . . w (n)] 0 1 L−1 murmurs (shown in Figure 1). is the weight vector at the time n that can be minimized, as /e lung sound auscultation is performed on the left shown in down posterior and right anterior regions of the chest. /e 2 2 (3) lung sounds considered in this study are described in Table 1. ‖w(n)‖ � ‖w(n + 1) · · · w(n)‖ . 2.2. ANC Algorithm Architecture. /e present study in- (ii) /e μ value for the input vector is calculated as vestigates on the performance of the combination of ANC follows: Amplitude Journal of Healthcare Engineering 3 Table 1: Lung sound data. N Name Type of sound Status Recording position SNR (db) in 1 LSN1 Vesicular Normal Posterior: left, low −8.65 2 LSA2 Crackles Abnormal Posterior: left, middle −14.4 3 LSN3 Bronchial Normal Chest: right, up −3.93 4 LSA4 Wheeze Abnormal Posterior: left, middle −15.9 5 LSN5 Broncho-vesicular Normal Posterior: left up −7.78 6 LSA6 Crackles Abnormal Posterior: right, low −53.8 Table 2: Summary of the NLMS algorithm. Tap-weight vector, w(n), Input: Input vector, x(n), and desired output, d(n) Output: Filter output, y(n), tap-weight vector update, w(n + 1) 1. Filtering output signal: y(n) � w (n)x(n) 2. Error estimation: e(n) � d(n) − y(n) 3. Tap weight and step size parameters adaptation: w(n + 1) � w(n) + α(x(n)/β + x(n) ) × e(n) α Original signal, X (n)= d(n)+ h (n) e(n) Output i 0 Lung μ � � � , � �2 (4) � � sound β + �X � n – where β is a small positive constant used to avoid division by zero when the input vector X is zero. /us, the problem on y(n) Heart Reference signal obtaining a gradient noise amplification in tap weights is ANC–NLMS sounds h (n) solved. Furthermore, α is the adaptation positive constant that is commonly less than 1 (0 < α < 1) [12, 13]. Figure 2: ANC-NLMS algorithm architecture. 2.3. NLMS Optimization. NLMS optimization is a principal approaches, that is, two parameters (α and β) will be method for minimal disturbance presented in [13], where controlled to satisfy the required performance. /e the error signal e (n) is defined as the difference between the i experiments are carried out with consideration of the desired signal and the filter output in (2). Hence, the error is following: minimized in magnitude and rearranged as follows: (a) /e adaptation constant α is changed within the e (n) � d (n) − Y (n) � d (n) − w (n + 1)x(n). (5) i i i i range of 0 < α < 1, and the small positive constant β is changed within the range 0.1–0.009 [12]. /e NLMS algorithm recursion obtains the constrained (b) /e VSS is studied within the range of 0-1 at the optimization criterion. /e tap weight is as follows: determined L value. x(n) (c) /e influence on the overall performance is w(n + 1) � w(n) + α × e(n). (6) β + x(n) monitored on the minimization of MSE, SNR out behavior, and algorithm output graphics. NLMS algorithm is an indication of the minimal dis- (ii) Second step: auto-optimum VSS turbance among iterations [13, 14]. Table 2 summarizes the NLMS algorithm. /e proposed idea here is a modified method from pseudorandom number generator μ for NLMS al- gorithm [16]. /e main parameters α and β are 2.4. NLMS μ . /e μ parameter should be optimized to n n changed randomly into variable value from random ensure the reliability of the designed algorithm [15] at 11 L numbers of distribution from 0 to 1 at each iteration values (determined previously) as a case study. /e optimal time. μ is obtained within a fixed optimal range of μ is obtained through the following steps. First, most ideal 0 ≥ μ � 0.1, which is explored experimentally from VSS is randomly searched. Results from the first step are the first step. /e proposed idea is implemented, as used in the second step. Such results include the imple- shown in Table 3; it achieves the optimum solution of mentation and automation of the algorithm work. Both steps NLMS in Section 2.3. are described in further detail in the following paragraphs. (i) First step: random search for the most ideal 2.5. Performance Analysis possible μ μ presents two main parameters, namely, α and β, 2.5.1. MSE. MSE is a performance function of AF, and its which can affect the overall performance of the al- target is the low MSE value for it to achieve a proper per- gorithm. /is aspect is the motivation for the VSS formance [13]. /erefore, the values and graph of this 4 Journal of Healthcare Engineering Table 3: Summary of optimum ANC-NLMS algorithm for HSS Table 4: Calculated MSE during searching for the optimal VSS. cancellation. MSE For time index, n � 1, 2,. . ., L filter length L (number of iteration) # µ L � 4 16 64 128 256 1024 and j � [2, 4, 8,. . ., 2048] 1 0.6 0.00003 Inf NaN NaN NaN NaN /e number of L value (j), 2 0.2 0.00210 0.00679 NaN NaN NaN NaN N (1, . . . , 4) the number of step size 3 0.1 0.00329 0.00329 Inf NaN NaN NaN Tap-weight vector, w(n), 4 0.09 0.00336 0.00590 2052.75 Inf NaN NaN Input Input vector, x(n) 5 0.041 0.00347 0.00500 0.00490 0.00864 Inf NaN Desired output, d(n) 6 0.011 0.00291 0.00155 0.00865 0.00423 0.00475 0.01501 Alpha � rand(1, N) 7 0.009 0.00278 0.00028 0.00150 0.00371 0.00413 0.01327 Beta � rand(1, N) Filter output, y(n) Output Tap-weight vector update, w(n + 1) 1. Filtering y(n) � w (n)x(n) E􏼂Y (n)􏼃 2. Error estimation e(n) � d(n) − y(n) SNR (dB) � 10 log 10 , (9) out µ ·L􏼁 n j For i � 0 : L − 1 E e (n) 􏼂 􏼃 For j � 0: N − 1 m(j) � mu/(x(n)^2 + be) where Y (n) is the output (pure LSS) of ANC-NLMS and If m(j) > mu max considered the signal, and e (n) is the estimated error (noise 3. Step size m(j) � mu max measurement) of ANC-NLMS and considered the noise. /e calculation If m(j) < mu min higher output SNR (SNR ) than that of SNR indicates the out in m(j) � mu min pureness of the obtained LSS and success of the noise removal End and consequently improves the performance of ANC-NLMS. End /e improvement level is estimated as follows: 4. Tap weight and step size 2 SNR (dB) � SNR − SNR . (10) imro. out in w(n + 1) � w(n) + μ(x(n)/β + x(n) ) × e(n) parameters adaptation 2.5.3. Output Graphics. Visual graphics are used as metrics in observing the change in input and output graphics. /ese quantity are essential to evaluate the performance of the AF. graphics will illustrate the input signals (original and ref- /e formula for MSE is given by the following equation: erence) in two windows and two other windows for output MSE(n) � E􏽮e (n)􏽯, (7) signals (pure LSS and estimated error). Accordingly, the change can be easily observed. where E[·] denotes the statistical expectation, and e is the /e experiment is carried out using MATLAB platform, estimated error of AF. /e MSE is calculated for the evo- in which an algorithm code is designed to obtain the main lution of AF performance during searching for the optimal output signals, their graphic matrix (SNR ) (11) SNR , and out in VSS, as shown in Table 4. other input parameters. SNR SNR SNR · · · SNR 􏼂 􏼃 􏼂 􏼃 􏼂 􏼃 􏼂 􏼃 out out out out 2.5.2. SNR Evaluation. SNR is used as a metric to estimate ⎢ ⎥ ⎡ ⎢ ⎤ ⎥ ⎢ ⎥ ⎢ μ (1),ℓ �2 μ (1),ℓ �4 μ (1),ℓ �8 μ (1),ℓ �2048 ⎥ ⎢ n j n j n j n j ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ the performance of the proposed method, and it is defined as ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 􏼂SNR 􏼃 􏼂SNR 􏼃 􏼂SNR 􏼃 · · · 􏼂SNR 􏼃 ⎥ ⎢ ⎥ ⎢ out out out out ⎥ the ratio of the amount of signal to the amount of noise [17]. ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ μ (2),ℓ �2 μ (2),ℓ �4 μ (2),ℓj�8 μ (2),ℓ �2048 ⎥ ⎢ ⎥ ⎢ n j n j n n j ⎥ ⎢ ⎥ ⎢ ⎥ In the present study, SNR is calculated before and after ⎢ ⎥ ⎢ ⎥ ⎢ ⎥, (11) ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ applying the ANC-NLMS algorithm to compare their values ⎢ 􏼂SNR 􏼃 􏼂SNR 􏼃 􏼂SNR 􏼃 · · · 􏼂SNR 􏼃 ⎥ ⎢ ⎥ ⎢ out out out out ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ μ (3),ℓ �2 μ (3),ℓ �4 μ (3),ℓ �8 μ (3),ℓ �2048 ⎥ ⎢ ⎥ ⎢ n j n j n j n j ⎥ for the same signals at the determining condition. /e input ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ SNR (SNR ) of the recording signal is measured in am- ⎢ ⎥ ⎢ ⎥ in ⎣ ⎦ SNR SNR SNR · · · SNR 􏼂 􏼃 􏼂 􏼃 􏼂 􏼃 􏼂 􏼃 out out out out plitudes; thus, SNR must be squared to be proportional to μ (4),ℓ �2 μ (4),ℓ �8 μ (4),ℓ �8 μ (4),ℓ �2048 in n j n j n j n j power, as expressed in (8) [18, 19]. SNR is used as one of metrics for the improvement of AF out E􏼂X (n)􏼃 performance during searching for the optimal VSS. /e (8) SNR (dB) � 10 log 10 , in E􏼂h (n)􏼃 obtained SNR is shown in Table 5. where X (n) is the original signal defined in (1) and con- sidered the signal, h (n) is the reference signal and considered 2.6. Experiment Procedures. /e experiment procedures are the noise, and i refers to the same number of pair signals. summarized as follows: Moreover, E(·) denotes operations in calculating the ex- (i) Create the coding program. pectation calculation in the time domain. According to the (ii) Unite the frequency sampling (8000 Hz). proposed method, (8) is suitable for SNR calculation because h (n) is correlated with existing noise (h (n)) in X (n). (iii) /e maximum duration of studied signal is 3.5 s, i 0 i /e output SNR value after applying ANC-NLMS is that is, one completed breathing cycle, which is out given by the following equation: equal to 28000 samples. Journal of Healthcare Engineering 5 Table 5: A sample of searching the optimal µ for the designed algorithm (SNR � −7.78). n in SNR (dB) out # µ L � 4 8 16 32 64 128 256 512 1024 2048 1 0.8 −5.3 −3.4 −1.8 −0.4 2.1 0.0 NaN NaN NaN NaN 2 0.6 −6.1 −4.1 −2.4 −0.9 1.4 0.0 NaN NaN NaN NaN 3 0.35 −8.0 −5.7 −3.8 −2.2 −0.1 3.0 0.0 NaN NaN NaN 4 0.26 −9.1 −6.6 −4.6 −2.9 −0.9 2.3 6.3 NaN NaN NaN 5 0.178 −10.7 −8.0 −5.7 −3.9 −2.0 1.1 5.5 7.5 NaN NaN 6 0.09 −13.5 −10.7 −8.1 −5.8 −3.9 −0.9 3.6 6.4 7.6 NaN 7 0.0797 −13.9 −11.1 −8.4 −6.2 −4.2 −1.2 3.3 6.2 7.5 0.0 8 0.0088 −24.5 −20.9 −17.6 −14.6 −11.8 −8.8 −5.0 −1.4 1.6 3.8 Learning curve for MSE –16 –17 –18 –19 –20 –21 –22 –23 –24 –25 –26 0 200 400 600 800 1000 1200 Number of iterations (n) mu = 0.0285 mu = 0.085 mu = 0.0334 mu = 0.063 mu = 0.0100 Figure 3: MSE performance of the NLMS for various optimal μ values and when L � 64. (iv) Experimentally identify the optimal μ range as were considered because of their effects on the performance stated in Section 2.5. of the designed algorithm. Figure 3 shows the MSE of μ with a value of 0.06, which (v) Procedure is performed with μ for each L value n 1 j displays faster convergence rate than those of others. Addi- (i.e., 11 times according to the j values) to calculate tionally, the AF became steady after approximately 200 it- and obtain the SNR 11 times at each L value. out j erations at steady state error of approximately −24 dB. /e (vi) /e procedure is repeated similarly with each μ other MSE tools needed a long time to converge and be- value, that is, four values within the determined came steady after approximately 400 iterations at steady optimal μ value, to obtain 44 total processing for state errors of approximately −26 dB for μ � 0.033 and signal at (μ · L ), where j � 2, 4, 8, 16, . . . , 2048. n j −25.5 dB for μ � 0.028 and 0.0085. /us, the steady state /erefore, SNR is calculated 44 times and updated out errors were small. with each μ to obtain the matrix shown in (11). ni Figure 4 illustrates the results for large VSS that results in (vii) /e experiment is carried out on MATLAB platform, unstable performance and unsatisfied results. According to in which an algorithm code is designed to obtain the the comparison between Figures 3 and 4, the performance main output signals and performance analysis tools. was good when VSS was small and close to zero. /e same conclusion was observed in the changes in MSE and SNR out values; they improved gradually with decreasing μ and 3. Results when they became close to zero, as shown in Tables 4 and 5. To obtain reliable results during all procedures, including the Searching for the optimal VSS identified the VSS optimal searching for the optimal VSS, the number of samples and L range of 0 ≥ μ � 0.1. /us, the designed algorithm lost its j n Mean (error ) (dB) 6 Journal of Healthcare Engineering Learning curve for MSE 0.02 0.018 0.016 0.014 0.012 0.01 0.008 0.006 0.004 0 200 400 600 800 1000 1200 Number of iterations (n) mu = 0.1261 mu = 0.1161 Figure 4: MSE performance of NLMS for μ > 0.100 values and L � 64. stability when μ was used without the identified range, as started from L � 64 and increased with increased L value. n j j shown in Tables 4 and 5 and Figures 3 and 4. /e improvement level based on SNR also increased with out Figures 5(a)–5(c) display that the maximum amplitude increased L value and obtained small change at different of input signal (Figure 5(a)) is 0.18, the maximum amplitude VSS. of the error signal is approximately 0.0015 (Figure 5(c)), and /e SNR values of normal LSS ranged from −3.93 dB to in the maximum amplitude of square error is 0.00004 (Figure 5(b)). −7.78 dB (Table 1). /e SNR values changed according to out /is result showed an increase in the algorithm number of the L and determined μ (Table 7). j n computation windows and minimization of error, as well Figure 7 exhibits the signal obtained by the designed as MSE, which is considered a function of NLMS perfor- algorithm, with the visual difference between the original mance. /erefore, NLMS optimization achieved minimal signal (graphic 1, Figure 7(a)) and the pure LSS (graphic 3, disturbance, and the designed algorithm accurately adapted Figure 7(c)). In general, the frequency components were low and converged to separate HSS from the original signal in the subjected original signal and decreased after applying (Section 2.3). AF on the pure LSS (graphic 3, Figure 7(c)). /e auto-optimal algorithm of identifying VSS has been used to evaluate the separation of HSS from the original 4. Discussions signal that consists of contaminated LSS and HSS based on SNR, as well as the performance of ANC and NLMS al- /e optimized results of the designed algorithm determined gorithm combination. the optimal VSS range of 0 ≥ μ � 0.1; in this range, the AF Table 6 shows the SNR values calculated from the AF became highly stable with nonstationary biosignals, such out outputs for the abnormal LSS case. Results of the com- LSS, where the performance of the proposed method showed parison of the SNR values shown in Table 1 and the SNR the most ideal trade-off between convergence speed and low in out values shown in Table 6 indicated that the SNR values were steady error on the basis of the appropriately autoselected µ in located from −14.4 dB to −53 dB in an abnormal case. /is [7, 15]. /is achievement approved the proper work of the result suggested that the abnormal LSS included high designed algorithm and its capability to separate signals by amount of noise, and the SNR values changed according identifying the VSS range on NLMS algorithm, which is in out to the L and determined VSS. agreement with the results of several works [20–22]. Figure 6 demonstrates the visual difference before and /e SNR values improved progressively at determined out after applying the proposed method at the determined pa- μ and L values for abnormal and normal lung sounds, as n j rameters where the original signal (graphic 1, Figure 6(a)) shown in Tables 6 and 7, respectively. /e SNR matrix out showed higher frequency components than that of pure LSS indicated that the performance level in the normal and (graphic 3, Figure 6(c)). In addition to the improved level of abnormal LSSs started improving from the L values of SNR , these results indicated the separation of the noise 64,128 and 256,512, respectively. Moreover, L can be out j�1024 components from the desired LSS. the upper limit at which the AF may work stably with normal Table 7 shows the improved performance level of sep- LSS and obtain accurate outputs. AF can work at L more arating signals on normal LSS case by observing SNR that than 2048 with abnormal LSS. /is result can be due to that it out Mean (error ) (dB) Journal of Healthcare Engineering 7 0.2 0.1 –0.1 –0.2 –0.3 0 0.5 1 1.5 2 2.5 ×10 (a) –4 ×10 0 0.5 1 1.5 2 2.5 ×10 Sample number (n) (b) –3 ×10 1.5 0.5 0 0.5 1 1.5 2 2.5 ×10 (c) Figure 5: Comparison of the amplitude of AF original signal X (n) and obtained errors. (a) Original signal. (b) MSE of the NLMS in the sample when μ � 0.09 and L � 128. (c). Output error of the NLMS. Table 6: SNR after applying the designed algorithm for an abnormal LSS case. L values LSS# µ 2 4 8 16 32 64 128 256 512 1024 2048 SNR (dB) out 0.111 −45.43 −38.31 −34.66 −16.16 −24.08 −19.57 −15.43 −15.09 −11.04 −5.06 −1.05 0.036 −45.43 −40.01 −32.09 −29.44 −13.39 −20.63 −16.99 −15.34 −3.65 −7.40 −2.16 LSA2 0.022 −28.12 −40.01 −32.75 −29.44 −24.66 −18.93 −13.87 −10.41 −3.41 −2.84 −3.08 0.016 −45.43 −40.01 −19.56 −19.31 −24.66 −17.19 −16.99 −15.34 −11.04 −7.95 −3.95 0.037 −35.28 −42.26 −40.03 −34.64 −16.15 −24.17 −17.89 −15.35 −14.04 −11.12 −3.80 0.028 −51.46 −45.66 −37.56 −34.64 −26.35 −24.40 −11.74 −13.11 −13.46 −11.12 −4.45 LSA4 0.025 −51.46 −44.18 −39.50 −31.80 −29.32 −13.80 −19.17 −18.29 −14.04 −3.80 −4.79 0.010 −37.72 −41.70 −40.03 −34.64 −16.15 −23.30 −20.41 −18.29 −6.95 −7.40 −8.00 0.100 −106.11 −101.49 −95.90 −83.13 −82.24 −79.34 −75.47 −56.85 −42.00 −43.41 −32.31 0.039 −101.38 −113.39 −89.27 −83.13 −86.10 −83.84 −65.40 −48.74 −41.56 −36.56 −40.06 LSA6 0.037 −101.38 −115.35 −109.27 −103.13 −89.80 −87.29 −82.06 −58.56 −57.56 −47.07 −40.58 0.017 −102.41 −95.35 −95.59 −100.87 −80.84 −70.46 −64.75 −48.74 −57.63 −36.11 −46.87 consists of normal LSS for low-frequency components, and between both LSS types whether in terms of SNR or SNR in out the abnormal LSS consists of high-frequency components. values. /e SNR values revealed several distinct markers /e performance estimation of ANC-NLMS algorithm out between LSSs; the normal LSS shows high SNR values, that combination based on automatic identification of the op- is, the SNR value reaches close to 0 or the positive axis. By timal VSS validated the correctness of the proposed contrast, the SNR values of the abnormal LSS are consid- method and its sequence steps in separating HSS from erably small in the negative axis. /erefore, these charac- LSS. Additionally, such estimation explored the distinct teristics may reveal clear difference that can differentiate features differentiating normal LSS from abnormal LSS, Mean (error ) 8 Journal of Healthcare Engineering –3 e original LSS corrupted by HSS ×10 –2 (a) e heart sound signal as reference –1 (b) e lung sound aer adaptive filter (Y) 0.2 –0.2 (c) e error –1 (d) Figure 6: Input and output signal graphics of ANC-NLMS algorithm on an abnormal case (LSA6). Obtained graphics and SNR � −40.58 out at μ � 0.039 and L � 2048, where the input SNR is −53.86 dB. Table 7: SNR after applying the designed algorithm on normal LSS. L values LSS# µ 2 4 8 16 32 64 128 256 512 1024 2048 SNR (dB) out 0.064 −24.82 −16.36 −16.82 −13.63 −11.06 −8.52 −6.18 −2.24 0.83 3.76 17.64 0.021 −24.82 −20.59 −16.82 −13.63 −11.06 −6.51 −7.00 −2.24 2.86 3.58 12.39 LSN1 0.017 −24.82 −20.59 −16.82 −8.11 −10.66 −8.18 −3.34 −0.93 3.52 3.73 11.25 0.010 −24.75 −20.59 −16.82 −11.11 −10.85 −9.20 −7.00 −2.24 3.46 3.58 8.83 0.025 −20.90 −18.84 −16.76 −14.79 −13.37 −9.60 −9.49 −8.26 −3.23 −3.68 −0.63 0.011 −21.17 −16.92 −16.76 −13.31 −12.31 −12.28 −9.49 −2.63 −4.94 −3.68 −1.64 LSN3 0.010 −21.17 −17.52 −16.76 −14.79 −13.42 −12.46 −4.64 −5.35 −5.81 −3.68 −1.42 0.009 −21.17 −18.84 −16.68 −14.79 −13.37 −11.90 −8.21 −2.68 −5.81 −3.68 −1.95 0.026 −25.95 −24.07 −19.72 −17.32 −9.96 −10.56 −8.47 −4.76 −0.14 1.78 6.56 0.020 −24.56 −13.45 −19.11 −12.52 −12.21 −10.51 −8.47 −4.76 2.15 5.02 6.02 LSN5 0.013 −27.94 −21.41 −18.37 −17.32 −14.24 −10.29 −8.47 −4.76 −0.78 1.78 4.88 0.010 −26.35 −19.43 −18.55 −17.32 −9.71 −8.15 −8.47 −4.76 −1.10 1.78 3.94 and these characteristic may be used as primary features to VSS for 11 L values to separate HSS from LSS. /e per- classify LSS. formance of the designed algorithm evaluated at determined conditions showed good result by reducing and minimizing the error gradually to zero after the convergence time. 5. Conclusions /e effectiveness of the designed algorithm to separate /is study investigated an effective method of ANC-NLMS HSS from contaminated LSS estimated based on the SNR out algorithm based on automatic identification of the optimal illustrated a progressive performance improvement level Journal of Healthcare Engineering 9 e original LSS corrupted by HSS 0.5 –0.5 (a) e heart sound signal as reference –1 (b) e lung sound aer adaptive filter (Y) –1 (c) e error –1 (d) Figure 7: Input and output signal graphics of ANC-NLMS algorithm on a normal case (LSN1). /e obtained graphics and SNR at out μ � 0.017, where the SNR is −8.65 dB, and the SNR is −3.34 dB at L � 128. n in out [3] K. Sathesh and N. J. Muniraj, “Real time heart and lung sound with increasing L and significantly improved separation of separation using adaptive line enhancer with NLMS,” Journal HSS from LSS. of ;eoretical and Applied Information Technology, vol. 65, /is SNR explored a novel method to differentiate out no. 2, pp. 559–564, 2014. between normal and abnormal LSSs. /is method may be [4] N. Q. Al Naggar and H. Ghazi, “Design two-channel in- used as basis in developing computerized diagnosis and strument to record lung and heart sounds at the same time, automating LSS calcification. and separate them using ANC-NLMS algorithm,” In- /e proposed approach clarified the correctness of the ternational Journal of Advanced Research in Electrical, Elec- combined designed algorithm and achieved significant tronics and Instrumentation Engineering, vol. 4, no. 4, performance. /e proposed method may be subject for pp. 2601–2609, 2015. further study on LSS under different settings and durations. [5] T. Tsalaile and S. Sanei, “Separation of heart sound signal from lung sound signal by adaptive line enhancement,” in Pro- Data Availability ceedings of the 15th European Signal Processing Conference, vol. 15, pp. 1231–1235, Poznan, Poland, 2007. /e data used to support the findings of this study are [6] N. Q. Al-Naggar, “A new method of lung sounds filtering available from the corresponding author upon request. using modulated least mean square adaptive noise cancella- tion,” Journal of Biomedical Science and Engineering, vol. 6, no. 9, pp. 869–876, 2013. Conflicts of Interest [7] R. Nersisson and M. M. Noel, “Hybrid nelder-mead search based optimal least mean square algorithms for heart and lung /e authors declare that they have no conflicts of interest. sound separation,” Engineering Science and Technology, an International Journal, vol. 20, no. 3, pp. 1054–1065, 2017. References [8] S. Sebastian and S. Rathnakara, “Separation of heart sound artifact from respiratory signals using singular spectrum based [1] S. Reichert, R. Gass, C. Brandt et al., “Analysis of respiratory advanced line enhancer,” International Journal on Advanced sounds: state of the art,” Clinical Medicine Circulatory, Re- Computer ;eory and Engineering, vol. 2, pp. 106–111, 2013. spiratory and Pulmonary Medicine, vol. 2, pp. 45–58, 2008. [9] M. Molaie, S. Jafari, M. H. Moradi, J. C. Sprott, and [2] R. K. /enua and S. K. Agarwal, “Simulation and performance analysis of adaptive filter in noise cancellation,” International S. M. R. H. Golpayegani, “A chaotic viewpoint on noise re- duction from respiratory sounds,” Biomedical Signal Pro- Journal of Engineering Science and Technology, vol. 9, pp. 4373–4378, 2010. cessing and Control, vol. 10, pp. 245–249, 2014. 10 Journal of Healthcare Engineering [10] O. El B’charri, R. Latif, K. Elmansouri, A. Abenaou, and W. Jenkal, “ECG signal performance de-noising assessment based on threshold tuning of dual-tree wavelet transform,” Biomedical Engineering Online, vol. 16, no. 1, 2017. [11] H. T. Ma and Y. T. Zhang, “Effects of the physiological pa- rameters on the signal-to-noise ratio of single myoelectric channel,” Journal of Neuroengineering and Rehabilitation, vol. 4, no. 1, p. 29, 2007. [12] K. Sathesh, N. J. R. Muniraj, A. V. Akshitha, B. K. Roy, M. Induja, and M. D. A. Maharasi, “Design and imple- mentation of real time electronic stethoscope with a method for separating heart sound from lung sound using TMS320C6713,” DSK Middle-East Journal of Scientific Re- search, vol. 23, no. 8, pp. 1909–1920, 2015. [13] B. Farhang-Boroujeny, Adaptive Filters ;eory and Applications, University of Utah, Wiley & Sons, Somerset, NJ, USA, 2nd edition, 2013. [14] S. Zhao, “Performance analysis and enhancements of adaptive algorithms and their applications of computer engineering,” Doctor of Philosophy thesis, Nanyang Technological Uni- versity, Singapore, 2009. [15] S. S. Haykin, Adaptive Filter ;eory, Pearson Education, New York, Ny, USA, 4th edition, 2008. [16] S. Jimaa, “Convergence evaluation of a random step-size NLMS adaptive algorithm in system identification and channel equalization,” in Adaptive Filtering, InTech, London, UK, 2011. [17] J. L. Semmlow, Biosignal and Biomedical Image Processing: MATLAB-Based Applications, Marcel Dekker, New York, NY, USA, 2004. [18] B. Y. Lu, “Unidirectional microphone based wireless recorder for the respiration sound,” Journal of Bioengineering and Biomedical Science, vol. 6, no. 3, p. 195, 2016. [19] A. H. Said, “Design of a chopper amplifier for use in bio- medical signal acquisition,” Master thesis, Department of Electrical and Computer Engineering, Southern Illinois University, Carbondale, IL, USA, 2010. [20] R. M. A. Falcão, “Adaptive filtering algorithms for noise cancellation,” Masters thesis, Faculdade de Engenharia da Universidade do Porto, Porto, Portugal, 2012. [21] S. A. Jimaa, N. Al Saeedi, S. Al-Araji, and R. M. Shubair, “Performance evaluation of random step-size NLMS in adaptive channel equalization,” in Proceedings of the 2008 1st IFIP Wireless Days, Dubai, UAE, November 2008. [22] K. Gomathi, V. Saravanan, and N. Santhiyakumari, “Variable step size for improving convergence of FxLMS,” Procedia Technology, vol. 25, pp. 420–426, 2016. International Journal of Advances in Rotating Machinery Multimedia Journal of The Scientific Journal of Engineering World Journal Sensors Hindawi Hindawi Publishing Corporation Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 http://www www.hindawi.com .hindawi.com V Volume 2018 olume 2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Journal of Control Science and Engineering Advances in Civil Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Submit your manuscripts at www.hindawi.com Journal of Journal of Electrical and Computer Robotics Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 VLSI Design Advances in OptoElectronics International Journal of Modelling & Aerospace International Journal of Simulation Navigation and in Engineering Engineering Observation Hindawi Hindawi Hindawi Hindawi Volume 2018 Volume 2018 Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com www.hindawi.com www.hindawi.com Volume 2018 International Journal of Active and Passive International Journal of Antennas and Advances in Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

Journal

Journal of Healthcare Engineering – Hindawi Publishing Corporation

Published: Jul 12, 2018

There are no references for this article.