Multi-channel speech enhancing method based on auditory perception model

Abstract

The invention relates to a multi-channel speech enhancing method based on an auditory perception model. The method comprises the steps that input signal non-even channels are divided into multiple channel signals; the noise level of each channel is detected, and noise level data are obtained; according to the noise level data, the channel gain of each channel is obtained by computing; the product of the channel signal and the channel gain of each channel is used as a gain signal of the channel; the gain signal of each channel subjected to signal synthesizing, and output signals are obtained; and the output signals are transmitted. A filter which simulates the auditory perception model and is used in the method combines a weighting splicing adding structure and all-pass conversion, the fact that human ear distinguishability is simulated under the situation of a small number of the channels is achieved, and meanwhile computation complexity is low. All-pass inverse transformation operation is added in the signal synthesizing process, the problem of phase distortion in the prior art is solved, and the method can be used for real-time signal processing.

Description

technical field [0001] The invention relates to a speech digital signal processing technology, in particular to a multi-channel speech enhancement method based on an auditory perception model. Background technique [0002] Speech enhancement is an important branch of speech signal processing, and its purpose is to improve sound quality, improve clarity and intelligibility, and reduce auditory fatigue. One of the main methods of speech enhancement is spectral subtraction, which estimates the power spectrum of clean speech by subtracting the noise power spectrum from the noisy speech power spectrum. The traditional spectral subtraction is to subtract a same spectral subtraction parameter from the entire frequency domain after a frame of speech undergoes fast Fourier transform. However, the non-stationary noise in the speech and the actual environment is non-uniformly distributed in the frequency domain. Correspondingly, noise signals have different influences on speech signa...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to address the defects of the prior art, provide a multi-channel speech enhancement method based on the auditory perception model, realize the simulation of human auditory resolution in the case of a small number of channels, the present invention has both weighted splicing The high efficiency of the summation structure overcomes the problems of real-time implementation and phase distortion in the current frequency conversion filter bank method

Images