The Research On Voice Signal Denoising Algorithm Based On Hilbert-Huang

Speech signal not only has the characteristics of containing the largest amount of information, but also is an important part in the field of speech signal processing. Speech signal in real life is non-stationary time-varying stochastic signal and will be affected by noise pollution, so we must do denoising processing to speech signal. For non-stationary random signal, Fourier transform, short-time Fourier transform and wavelet transform processing result is not very good, the Hilbert Huang transform has the multi-resolution and highly adaptive feature makes it very suitable for processing non-stationary nonlinear time-varying stochastic signal, this paper presented the research on voice signal denoising algorithm based on Hilbert-Huang.First of all, this paper introduced the basic features of speech and noise, the basic theory of the Fourier transform, short-time Fourier transform and the wavelet transform, the denoising principles of several common speech signal denoising method, and the commonly used speech signal quality evaluation standard.Secondly, this paper detailed elaborated the basic theory and the algorithm's implementation processing of Hilbert-Huang transform, and analyzed the Hilbert-Huang transform parsing process of the EMD decomposition and the Hilbert transform, and simulation to achieve the three different kinds of signals of the Hilbert Huang transform.Thirdly, according to the short-time stationarity of speech signal, the cubic spline interpolation method in the process of curve fitting average envelope signal occurring undershoot and overshoot phenomenon and cut-off selection screening problem according to noise and useful signal's different energy distribution of IMF component in the process of Hilbert Huang transform, this paper put forward the improvement of the Hilbert Huang speech signal denoising algorithm.Finally, using matlab for simulation comparison, respectively using wavelet transform and Hilbert Huang transform and the improved Hilbert Huang transform in this paper to deal with speech signal which added noise, the simulation results showed that: in this paper, the improved Hilbert Huang algorithm has better denoising effect, the speech signal after denoising not only has good time-frequency waveform, but also has high signal-to-noise ratio.