Wavelet Decomposition using Matching Wavelet Function for Feature Extraction In EEG based BCI
Wavelet transform acts as the powerful tool for feature extraction as it map the wavelet with the signal and take out the required variations from the signals. This work empirically selected the matching wavelets db10 and bior6.8 for signal decomposition. The obtained wavelet coefficients are used for preparing statistical and higher order statistical (HoS) features. HoS features are preferred by this research as it represents dynamics of the signal. Support Vector Machine used for classification acts as the robust classifier for BCI application. The work analyzes various wavelets functions and different kernel functions using the performance parameters resulting in 92% classification accuracy.