Audio segmentation based on wavelet transform