Voice activity detection method, voice activity detection apparatus and electronic device

Abstract

The invention provides a voice activity detection method, a voice activity detection apparatus and an electronic device. The voice activity detection method comprises: acquiring sound data; extractingenergy value data in the sound data; extracting a second feature in the sound data; judging whether the sound data meet a first preset condition or not according to the energy value data in the sounddata; and judging whether the sound data are voice data or not according to the second feature after the first preset condition is met. Compared with the prior art, the voice activity detection method of the invention achieves higher accuracy by judging twice. The invention can not only realize a detection effect of a traditional solution, but also detect whether the voice is human voice or not and whether the voice is environmental thermal noise upon encountering noise with large energy value, noise with long duration or sporadic noise. In addition, the solution of the invention increases the voice activity detection efficiency by 30%, and improves accuracy and practicability of voice activity detection.

Description

technical field [0001] The present invention relates to a voice activity detection technology, in particular to a voice activity detection method, a voice activity detection device, and electronic equipment using the voice activity detection method. Background technique [0002] VAD (Voice Activity Detection, Voice Activity Detection) is an important technology in intelligent voice interaction. In the intelligent voice interaction, after the active voice is detected, the voice is converted into text backwards, and finally the text is sent to the intelligent engine for processing. In intelligent voice processing, VAD mainly plays two roles: [0003] 1) After the human voice is detected, it is sent to ASR (Automatic Speech Recognition, automatic speech recognition) for processing, which can greatly save bandwidth and save the cost of using ASR; [0004] 2) ASR misrecognition caused by environmental noise can be avoided, which in turn leads to wrong processing actions of the ...

AI Technical Summary

Problems solved by technology

For intelligent voice interaction, it is necessary to highlight the detection of whether it is a human voice. These two solutions have many shortcomings and cannot meet this requirement.

Images