AC/V Segmentation Approach for Mandarin Speech Based on Wavelet Transform