A microphone-based voice activity detection system for earphone wearers

Abstract

The invention discloses a voice activity detection system for earphone wearers based on microphone technology, comprising: a microphone array, a first estimation module, a second estimation module and a joint control module; the microphone array is used for receiving sound signals; the first estimation module , used to determine the wearer’s first voice existence probability according to the incoming wave direction of the sound source; the second estimation module is used to determine the wearer’s second voice existence probability according to the direct reverberation ratio of the sound source; the joint control module , for determining a third voice presence probability according to the first voice presence probability and the second voice presence probability, and performing voice activity detection on the wearer. Using microphone array technology, it detects the voice activity of the earphone wearer. Even in complex acoustic scenarios such as low signal-to-noise ratio, high reverberation, and multi-speaker interference, it can detect the wearer's voice activity, providing an important basis for subsequent voice enhancement and voice recognition technologies.

Description

technical field [0001] The invention relates to the field of voice activity detection, in particular to a microphone technology-based voice activity detection system for earphone wearers. Background technique [0002] At present, earphones have become an indispensable electronic product in daily entertainment and voice communication. With the development of artificial intelligence technology, earphones are no longer limited to ordinary call functions, and the addition of functions such as voice enhancement and voice assistants have greatly improved the practicability of earphones. In practical applications, the microphone of the headset often picks up a variety of noises. The types of noise mainly include relatively stable environmental noises such as subway noise and road noise, and strong noises such as other nearby speakers or broadcasts from shopping malls and high-speed rail stations. non-stationary noise. These noises not only affect call quality, but also increase f...

AI Technical Summary

Problems solved by technology

[0003] For wearer voice activity detection (Voice activity detection, VAD), the existing methods are mainly divided into three categories. The first category is to use the voice enhancement algorithm based on the microphone array to enhance the voice in the target direction. [1，2] , using beamforming technology to enhance the wearer's voice and suppress interference from other directions, but when the signal-to-noise ratio is low, especially when there is strong interference voice, this type of method often fails to detect the target voice or mistakenly detects strong interference Speech as target speech, which affects beamforming performance

The second category is noise reduction based on auxiliary sensor methods [3-5] , use such as vibration sensors or acceleration sensors (such as bone conduction sensors) to detect the wearer's voice activity. This type of method requires auxiliary sensors to provide non-acoustic information for auxiliary judgment. Although the accuracy is high, the wearer needs to wear the headset correctly and higher cost

This method is not robust enough in practical application, and when the interfering voice is close to the target, it will lead to misjudgment

Images