Non-dyadic wavelet decomposition for sensory-motor imagery EEG classification
Abstract
Identification of imagined motor movement during cerebral activity is a significant task for translating the activity into control signals for brain–computer interface-based applications. This paper proposes a model based on non-dyadic wavelet decomposition for specifying left-hand and right-hand movement detection using motor imagery EEG signals. Three key components define our model: (1) The preprocessing and non-dyadic wavelet decomposition of EEG signals, (2) feature extraction using the common spatial pattern (CSP) coefficients of wavelet decomposed signals, (3) classification of test signal using selected features. The wavelet decomposition is done on the basis of m-band filtering. Classification of extracted features is done using different classifiers and obtained results are compared in terms of sensitivity, specificity, accuracy, and the kappa value. The proposed model gives the highest classification accuracy of 85.6% using decision tree classifier for BCI Competition 4 dataset IIa and BCI Competition 3 dataset IVa.