Voice denoising method and device, electronic device and storage medium

Abstract

The invention provides a voice denoising method and device, an electronic device and a storage medium. The method comprises the steps of obtaining to-be-denoised voice data; extracting spectrogram information of the to-be-denoised voice data; inputting the spectrogram information into a pre-trained neural network model to obtain a signal-to-noise ratio corresponding to the spectrogram information,the neural network model being obtained by training based on sample spectrogram information of sample voice data marked with known noise data and a sample signal-to-noise ratio corresponding to the sample spectrogram information; and denoising the to-be-denoised voice data based on the signal-to-noise ratio corresponding to the spectrogram information to obtain denoised voice data. When the neural network model is trained, the noise data in the adopted sample voice data is known, so that the trained neural network model can accurately determine the signal-to-noise ratio corresponding to the spectrogram information of the to-be-denoised voice data, and has very strong robustness in different voice environments, and thus the voice denoising effect can be greatly improved.

Description

technical field [0001] The present application relates to the technical field of speech recognition, and in particular to a speech denoising method, device, electronic equipment and storage medium. Background technique [0002] With the continuous development of technology, speech recognition has entered every aspect of people's lives, such as voice assistants and smart speakers. However, due to the existence of noise in the environment, the accuracy of speech recognition is seriously affected. In order to better improve the performance of speech recognition, it is necessary to have a corresponding speech denoising method to support it. [0003] The voice denoising method in the related art is generally based on the ratio of the power of each frequency band in the voice to the noise power to calculate the noise suppression factor of each frequency band, so as to multiply the amplitude of each frequency band by the corresponding noise suppression factor, so as to achieve voi...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the technical problems existing in related technologies that it is impossible to accurately determine what types of noise a piece of speech contains, as well as the frequency spectrum of the noise and the position of the noise in the speech, resulting in inaccurate calculation of the signal-to-noise ratio, which seriously affects the effect of speech denoising.

Images