Speech Pitch Period Detection Algorithm Based on Wavelet Transform and Spacial Correlation Function

The pitch period detection methods for noisy speech signal were analyzed. A pitch detection method for noisy speech signals based on wavelet transform and spacial correlation function was proposed. Pitch period of speech was detected based on the principle of singularity detection of signals with wavelet transform at different scales. And the spacial correlation function of wavelet transform was proposed to improve the detecting-accuracy and robustness. The spacial correlation coefficients between adjacent scales of wavelet transform coefficients were calculated after wavelet transform of noisy speech signal. And then the pitch period was detected by using the correlation coefficients to find the distance between the local maximums. The experimental results show that the pitch period extracted by this method has better robustness and accuracy compared with traditionary wavelet transform method.