A multiresolution approach to spike detection in EEG

A technique is proposed for the automatic detection of spikes in electroencephalograms (EEG). A multiresolution approach and a non-linear energy operator are exploited. The signal on each EEG channel is decomposed into three subbands using a non-decimated wavelet transform. Each subband is analyzed by using a non-linear energy operator, in order to detect, peaks. A decision rule detects the presence of spikes in the EEG, relying upon the energy of the three Subbands. The effectiveness of the proposed technique was confirmed by analyzing both test signals and EEG layouts.