Adaptive Segmentation Using Wavelet Transform

In many applications analysis of nonstationary signals requires the signal to be segmented into piece-wise stationary epochs. Segmentation can be performed by splitting the signal at time instants of charge in the amplitude or frequency content of the signal. In this paper, the signal is decomposed into signals with different frequency bands using wavelet transform. The nonlinear energy operator is then applied on the decomposed signals, which combines the amplitude and frequency contents of the signal. The proposed technique is applied on synthetic signal and real EEG data to evaluate its performance on segmenting nonstationary signals. The results show that the proposed technique outperforms the recently published method in decomposing nonstationary signals.