Mute detection method based on speech characteristic to jude

Abstract

The invention discloses a voice characteristic identification-based silence detecting method, firstly extracting multi-threshold overzero rate of an audio data frame; pre-identify silence by weighting the multi-threshold overzero rate to identify obvious silence; extracting composite characteristic of an audio data frame, where the composite characteristic comprises overzero rate, short-time energy value, and variable resolution frequency spectrum-based Mel scale revere spectrum coefficient; using dichotomy support vector machine to identify the composite character of the audio frequency, one class normal voice and the other class silence. And the invention can raise success rate of silence detection and can identify some special voices, able to be widely applied to network voice talking, especially voice chatting and video meeting.

Description

1. Technical field [0001] The invention relates to an audio processing method, in particular to a method for detecting silence based on speech feature discrimination for Internet voice calls. 2. Background technology [0002] In the process of human speaking, its voice can be divided into two parts: silence and voice, and 60% of the time is silence on average. And when many people are talking, basically only one person speaks at each moment, while the others are silent. Silence and noise (including airflow noise) introduced by voice collection equipment are transmitted in the network like voice data, causing voice quality to degrade. Using the mute suppression technology, the mute part can be eliminated, which can save more than 50% of the transmission bandwidth and reduce network congestion. [0003] Existing silence detection methods include extracting audio signal eigenvalues ​​and comparing them with preset thresholds to determine silence. The parameters used in tradit...

AI Technical Summary

Problems solved by technology

[0003] Existing silence detection methods include extracting audio signal eigenvalues ​​and comparing them with preset thresholds to determine silence. The parameters used in traditional silence detection methods include short-term zero-crossing rate, short-term energy, autocorrelation coefficient, but Speech signals and some background noise signals are non-stationary, so the system recognition rate is poor; moreover, because the threshold value is fixed, it cannot adapt to different noises well, so the recognition rate of these detection systems is not high

[0004] In addition, with the popularization of Internet voice calls, most of the applications are concentrated on the personal computer platform. For the convenience of use, the speaker generally chooses to wear a headset to communicate, which makes the microphone very close to the nose and mouth of the person. The airflow from your breath enters the microphone and produces an audio stream

Although this audio signal is relatively weak, it is also a kind of speech, and some commonly used silence detection methods (such as G.729B, G.723.1A, etc.) will recognize part of the airflow noise as normal speech, further reducing the detection system. Recognition rate

Images