Speech enhancement method of multi-band spectral subtraction based on adaptive spectrum base optimization

Abstract

The invention discloses a speech enhancement method of multi-band spectral subtraction based on adaptive spectrum base optimization, and aims to solve the problem that residual noise control is not ideal in an existing speech enhancement method adopting over-reduction spectral subtraction or multi-band spectral subtraction. On the basis of multi-band spectral subtraction, a spectrum bottom optimization function changing along with a signal-to-noise ratio is designed to fill spectrum valleys in different degrees under different noise conditions, and the method specifically comprises the following steps: firstly, smoothing noises estimated by a voice activity detector to reduce fluctuation of noise estimation; secondly, performing related adjustment on a sub-band fine adjustment factor according to an experiment to improve the accuracy of noise estimation, and finally filling the spectrum valley of the enhanced speech through a self-adaptive spectrum bottom optimization function to reduce a narrow-band noise peak value and music noise generated by spectrum over-reduction. The objective of the invention is to control the amount of residual noise under different signal-to-noise ratios to a low level so as to reduce music noise caused by a spectral subtraction process, thereby improving voice quality.

Description

technical field [0001] The invention belongs to the field of speech enhancement, in particular to a speech enhancement method based on multi-band spectrum subtraction based on adaptive spectrum floor optimization. Background technique [0002] Speech is an information carrier for human beings to communicate with each other. In the actual environment, the voice signal is often interfered by the external noise environment, which leads to the deterioration of the voice signal. Speech enhancement is a method to solve noise interference, which can extract the purest possible original language from noisy speech. There are two main purposes of speech enhancement: one is to improve speech quality and eliminate background noise, so that listeners are willing to accept it without feeling tired; the other is to improve speech intelligibility and make it easier for listeners to understand. [0003] Because there are many kinds of noises and their characteristics are not exactly the sa...

AI Technical Summary

Problems solved by technology

[0010] In order to overcome the existing oversubtraction spectral subtraction method, the oversubtraction factor and the lower limit factor of the spectrum are used in a stable noise environment to reduce the music noise, but the noise in the real environment is constantly changing, and different interference noises have different effects on each frequency band of speech. All the same, the effect is not ideal; the existing band spectrum subtraction uses a fixed spectrum lower limit factor without controlling the residual noise to a lower level, and the same effect is not ideal. The present invention designs a multi-band spectrum subtraction on the basis of The spectral bottom optimization function that varies with the signal-to-noise ratio fills the spectral valleys in different degrees under different noise conditions, aiming to control the amount of residual noise under different signal-to-noise ratios to a lower level, so as to reduce the noise caused by the spectral subtraction process. In order to improve the speech quality, a speech enhancement method based on multi-band spectral subtraction based on adaptive spectral floor optimization is proposed

Images