Analysis of non-stationary electroencephalogram using the wavelet transformation

Wavelet packet analysis is employed in this paper to investigate the transient characteristics of the practical electroencephalogram (EEG) signals. Since the non-stationary nature of different kinds of clinical EEG rhythms covers different frequency bands, wavelet packet decomposition, as a time-varying filter, can be used for forming the filters with different frequency response characteristic to detect 4 different EEG rhythms. We also select he coefficients of wavelet transformation corresponding to the desired rhythms bands to construct the transient brain topographic mapping with multi-channel signals. Several clinical EEG signals are decomposed into 4 kinds of rhythms, and the specified rhythms are investigated and compared so that we can further understand the dynamic rhythms of the EEG signals in different functional states of brain. It is indicated from the experimental results that the transient rhythms and the dynamic information of the brain electrical activities can be well described by using wavelet packet analysis. The method presented in this paper also proposes a new way for the analysis of other pathological EEG signals, such as the signals containing the spikes and slow waves.