Audio block processing method applied to intelligent voice mouse

Abstract

The invention discloses an audio block processing method applied to an intelligent voice mouse, which comprises the steps that framing processing is performed on each audio block of an audio fragmentacquired by the intelligent voice mouse; an average energy value of each audio block is calculated and compared with a preset threshold, wherein the audio block is a valid audio block if the average energy value is greater than or equal to the preset threshold, and the audio block is a mute audio block if the average energy value is less than the preset threshold; and the intelligent voice mouse uploads the valid audio blocks and judges whether the mute audio blocks conform to discarding conditions or not, the mute audio blocks, which do not conform to the discarding conditions, are uploaded,and the mute audio blocks conforming to the discarding conditions are discarded. According to the invention, the invalid mute audio blocks are eliminated, and only the valid audio blocks and the muteaudio blocks playing a transitional function are uploaded, thereby not only reducing the network transmission burden, but also reducing the burden of voice recognition, and greatly improving the efficiency of voice recognition.

Description

technical field [0001] The invention belongs to the technical field of voice recognition, and relates to an audio block processing method, in particular to an audio block processing method applied to an intelligent voice mouse. Background technique [0002] When the smart voice mouse works, a large number of audio blocks need to be uploaded to the Xunfei speech recognition server for speech recognition, which contains invalid audio blocks, which increases the burden on the speech recognition server and network transmission. [0003] The existing one is VAD detection, the main purpose is to detect the front and rear endpoints of the effective voice of the audio, and then perform speech recognition after obtaining the endpoints. VAD detection only detects the endpoints of the voice, and the audio does not reduce the burden on network transmission, and the voice endpoint detection exceeds a few seconds of silence After that, the recognition is automatically terminated, which se...

AI Technical Summary

Problems solved by technology

[0003] The existing one is VAD detection, the main purpose is to detect the front and rear endpoints of the effective voice of the audio, and then perform speech recognition after obtaining the endpoints. VAD detection only detects the endpoints of the voice, the audio does not reduce the burden on the network transmission, and the voice endpoint detection exceeds a few seconds of silence After that, the recognition is automatically terminated, which seriously affects the efficiency and effect of speech recognition.

Images