Optimal Design of Three-Band Orthogonal Wavelet Filter Bank with Stopband Energy for Identification of Epileptic Seizure EEG Signals

Abstract

We design three-band orthogonal wavelet filter bank using unconstrained minimization of stopband energies of low-pass, band-pass, and high-pass filters. The analysis polyphase matrix of the orthogonal filter bank is represented by the parameterized structures such that the regularity condition is satisfied by the designed perfect reconstruction filter bank (PRFB). Dyadic and householder factorization of the analysis polyphase matrix is employed to impose perfect reconstruction, orthogonality, and regularity order of one. Three-band orthonormal scaling and wavelet functions are generated by the cascade iterations of the regular low-pass, band-pass, and high-pass filters. The designed three-band orthogonal filter bank of length 15 is used for feature extraction and classification of seizure and seizure-free electroencephalogram (EEG) signals. The classification accuracy of 99.33% is obtained from the designed filter bank which is better than the most of the recently reported results.