Tree structured filter bank for time-frequency decomposition of EEG signals

A simple tree structured multirate filter bank geometry specifically tailored for EEG time-frequency analysis is presented. The whole structure is based on only one IIR allpass filter. It is computationally efficient and can be easily adaptable to different sampling rates. Octave band subband decomposition of EEG signals is obtained with final subbands in the filter bank tree closely approximating domains of characteristic EEG rhythms. The proposed structure is useful for EEG compression through subband coding and EEG rhythm identification in real time based on energy distribution in subbands.