A method is proposed for detecting time-varying rhythms of nonstationary el
ectroencephalograms (EEG). Multi-resolution decomposition is used to invest
igate the transition of clinical EEG signals. Wavelet packet transformation
is applied to design the filters with different frequency characteristics
in order to extract different kinds of dynamic EEG rhythms. Several actual
EEG signals with different brain function states are tested and analysed. T
he parameters of the wavelet packet transform corresponding to the rhythms
are developed to reconstruct the time-varying electrical brain activity map
ping. From the experimental results, the dynamic characteristics of clinica
l bl ain electrical activities can be demonstrated by using wavelet packet
decomposition. The presented method can be used for the analysis of other b
iomedical signals.