Analysis of Noise-Corrupted Speech Characteristics Based on Hilbert-Huang Transform

To improve the accuracy of speech/nonspeech decision by using Hilbert-Huang transform method to analyze the features of speech signals. Applying the Hilbert-Huang transform on the speech signals, the energy distribution of speech signals on time domain and frequency domain are obtained firstly, then,building the three-dimensional Hilbert-Spectrum of time-frequency-amplitude and analyzing the marginal spectrum. Finally validating the efficiency of speech processing based on Hilbert-Huang transform by the experiments of speech/nonspeech decision. The results of speech endpoint detection show that, after analyzing and denoising by Hilbert-Huang transform, the accuracy of detection improves markedly. Hilbert-Huang method gives the true description of the non-linear and non-stationary characteristics of speech signals, it has wide application prospect.