It is the property of a discrete time filter or a function which ensure minimizes the spreads or the variance in time and frequency domains.
Published in Chapter:
Three Channel Wavelet Filter Banks With Minimal Time Frequency Spread for Classification of Seizure-Free and Seizure EEG Signals
Dinesh Bhati (Acropolis Institute of Technology and Research, Indore, India), Akruti Raikwar (Acropolis Institute of Technology and Research, Indore, India), Ram Bilas Pachori (Indian Institute of Technology Indore, Indore, India), and Vikram M. Gadre (Indian Institute of Technology Bombay, Bombay, India)
Copyright: © 2020
|Pages: 17
DOI: 10.4018/978-1-7998-2120-5.ch012
Abstract
The authors compute the classification accuracy of minimal time-frequency spread wavelet filter bank with three channels in discriminating seizure-free and seizure electroencephalogram (EEG) signals. Wavelet filter bank with three channels generates two wavelet functions and one scaling function at the first level of wavelet decomposition. A time-frequency localized filter bank can be generated by minimizing the time spread and frequency spread of any one or all the functions simultaneously. The minimal time-frequency spread wavelet filter bank with three channels of regularity order, one designed with several different time-frequency optimality criteria and length six, are chosen, and the effect of each optimality criterion on the discrimination of seizure-free and seizure EEG signals is computed. The classification accuracy for five different optimality criteria are computed. Time-frequency localized three-band filter bank of length six classifies, the seizure-free and seizure EEG signals of Bonn University EEG database, with 98.25% of accuracy.