Sound processing method, device and equipment

Abstract

The embodiment of the invention provides a sound processing method, device and equipment. The method comprises the following steps that at least two feature vectors of an audio frame are obtained, wherein the audio frame comprises voice signals and noise signals, and acoustic characteristic type of the at least two feature vectors are different; fusion processing is conducted to the at least two feature vectors to obtain a fusion feature vector; the label value of the audio frame is obtained based on the fusion feature vector, wherein the label value is used for indicating the power ratio of the voice signals to the noise signals in the audio frame; and processing results of the audio frame are determined based on the label value of the audio frame and the noise signal power of the audio frame. The sound processing method, device and equipment are used for improving the effect of noise suppression.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of sound, and in particular, to a sound processing method, device, and equipment. Background technique [0002] With the continuous development of sound technology, in various application scenarios, noise signals need to be suppressed during sound suppression (including voice signals and noise signals), that is, noise reduction processing is performed on sounds. [0003] In the prior art, when noise reduction processing is performed on a sound, each audio frame in a piece of sound is processed separately. For each audio frame, the feature vector of the audio frame is obtained first, and the feature vector is processed by a preset model, so as to suppress the noise signal in the audio frame. However, in the actual application process, due to the existence of various types of noise in the environment, the audio frame may also contain multiple types of noise, and the feature vector of...

AI Technical Summary

Problems solved by technology

However, in the actual application process, due to the existence of various types of noise in the environment, the audio frame may also contain multiple types of noise, and the feature vector of the audio frame cannot represent the characteristics of all noises, so that according to the preset model When processing the feature vector, the noise in the audio frame cannot be accurately suppressed according to the feature vector of the audio frame

For example, the ability to suppress noise is low, or the speech is suppressed while suppressing the noise, or the speech is damaged while suppressing the noise

It can be seen from the above that in the prior art, the noise suppression effect is poor

Images