Audio signal processing method, device and storage medium

Abstract

The present disclosure relates to an audio signal processing method, device and storage medium. The present disclosure acquires a first reference audio signal and a mixed audio signal to be processed, wherein the mixed audio signal includes an echo signal and a target audio signal, and according to the mixed audio signal to be processed Signal and the first reference audio signal, use the pre-trained deep clustering network model, voiceprint recognition network model and multi-head attention mechanism model to eliminate the echo signal to obtain the target audio signal, which can eliminate the echo in the mixed audio in complex environments And effectively suppress residual echo, improve audio quality and audio processing performance, and increase user experience.

Description

Technical field [0001] The present disclosure relates to the field of computer applications, and more particularly to an audio signal processing method, a device, and a storage medium. Background technique [0002] Echo Eliminate the earliest application in the audio call system, at both ends of the call, the sound of one end passes through the line to the other end, and play out through the other end, the microphone of the other end receives the sound of the speaker, at the same time due to The orientation and reflection of the floor, wall, other objects, the microphone receives various reflection sounds in addition to the direct sound of the speaker play, this mixed sound will pass back to the end of the speech, there is an echo problem . The existence of echo interferes with people's conversations, reducing the quality of audio, and poor call effects. In the smart voice device, the audio played by the device itself will be received by its own microphone, and there is also an e...

AI Technical Summary

Problems solved by technology

This method can eliminate linear echo well, but due to nonlinear echo and time delay estimation error will introduce residual echo to a certain extent, although nonlinear processing can suppress this residual echo to a certain extent, but the degree of suppression is limited, There are still residual echoes, especially those in complex environments. The filter cannot quickly track the changes in the room’s impulse response, resulting in poor final echo cancellation. The echo suppression performance is average, the speech recognition rate is low, and the user’s listening experience Difference

Images