关键词: 语音信号, 复子波, 相位谱, 基音周期, 共振峰

Abstract: Chinese speech signal has characteristic of strong variability and the unstability.This paper proposes variable time-frequency complex wavelets method to pick-up information about Chinese speech signal from the wavelet analysis of amplitude spectrum，phase spectrum，pitch and formant information.This paper choose complex Gaussian wavelets which have the numbers of vanishing moments，and have better time-frequency character for extracting Chinese speech signal and phase.The experimental results show that the amplitude spectrum，phase spectrum，wavelet spectra of speech signal great describe the Chinese phonetic syllables envelope，envelope details and tone，extract the dynamic pitch period and accurately estimate resonance peaks.It provides a new way of extracting feature information and recognizing Chinese speech signal.

Key words: speech signal, complex wavelets, phase spectrum, pitch period, resonance peaks