Voice processing method and device thereof, electronic equipment and computer storage medium

Abstract

The invention discloses a voice processing method. The method comprises the following steps: constructing a training pair of first voice data and second voice data; respectively inputting the original features of the first voice data and the second voice data into a generation model, and generating first embedded data corresponding to the first voice data and second embedded data corresponding to the second voice data; inputting the first embedded data and the second embedded data into a discrimination model for training to obtain a discrimination result; according to the discrimination model, performing adversarial model learning in a stochastic gradient descent mode to obtain a voice noise reduction model; and performing noise reduction processing on the target voice data according to the voice noise reduction model. According to the method, on the premise that robustness is guaranteed and dependence on priori knowledge is small, relevance and difference between normal signals and noise signals are fully learned in a self-adaptive mode through the deep learning network, and a good voice noise reduction effect is achieved.

Description

Technical field [0001] The present invention relates to the field of speech processing, and more particularly to a speech processing method, apparatus, electronic device, and computer storage medium. Background technique [0002] With the development of voice communication systems, speech has entered all aspects, such as mobile phone audio and video calls, car calls, etc., the external environment usually affects the acupuncture and clarity of voice, and will cause audible fatigue to the listeners. The traditional speech noise reduction method is based on various signal filtration algorithms, such as Kalman filter algorithm, constructing a linear combination model of normal signal and noise signal to achieve noise separation, and the existence is that it cannot be deeply removed. Mining the correlation and difference between the normal signal and the noise signal, thereby affecting the noise separation effect, and requires the robustness of the model constructor to ensure the rob...

AI Technical Summary

Problems solved by technology

Traditional speech noise reduction methods are based on various signal filtering algorithms to build models, such as the Kalman filter algorithm, which constructs a linear combination model of normal signals and noise signals to achieve the purpose of noise separation. Mining the correlation and difference between the normal signal and the noise signal, which affects the noise separation effect, and requires the model builder to add certain prior knowledge to ensure the robustness of the model

Images