Improvement of pitch detection in noisy speech based on wavelet transform

An improved pitch detection algorithm in noisy speech based on wavelet transform is presented in this paper. The original speech signals are filtered by LPC inverse filter to obtain predictive residuals. The predictive residuals are decomposed by dyadic wavelet transform with several scales respectively. And the spatially selective noise filtration technique is applied to the decomposed signals to sharpen and enhance the glottal closure instants of voiced speech. The pitch periods are obtained from the enhanced signals. The simulations indicated that the performance of the proposed algorithm is improved considerably both for clean speeches and noisy speeches.