Audio signal processing method, model training method and related device

Abstract

The invention discloses an audio signal processing method, a model training method and a related device, and the method comprises the steps: inputting an audio input signal into a machine learning model, so as to obtain a first squeal point and a corresponding first gain value; and processing a second howling point in the audio input signal according to the second gain value to obtain an audio output signal. Therefore, howling in the audio input signal is suppressed, howling is suppressed in the initial stage, and gain circulation cannot be carried out. Due to the correspondence of the indication howling points in the machine learning model and the convenience of model calculation, the processing process is rapid, effective audio signals are not affected, and the accuracy and efficiency ofaudio processing are improved.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to an audio signal processing method, a model training method and related devices. Background technique [0002] With the development of mobile terminal-related technologies, more and more smart devices appear in people's lives. Among them, voice calls through smart devices are particularly prominent. However, during the call, the local microphone will also collect These audio signals may circulate during the speech process between the local end and the peer end, especially in the short-distance speech process, the audio signal will continue to loop gain, resulting in howling. [0003] Generally, a frequency shifter or a phase shifter can be used to process the input audio at the local end, that is, to destroy the audio information that is in phase with howling, so as to suppress howling. [0004] However, the method of using a frequency shifter or a phase shifter takes...

AI Technical Summary

Problems solved by technology

[0004] However, the method of using a frequency shifter or a phase shifter takes a long time to process, and is not suitable for real-time processing of voice calls, and due to the fixed phase removal of the howling point, it will also damage the sound quality of the effective audio. Affects the accuracy of audio processing

Images