Voice equipment DOA estimation enhancement method and device

Abstract

The invention provides a voice equipment DOA estimation enhancement method comprising the steps that the microphone signal when the voice equipment is awakened is acquired, and the voice awakening confidence of each frame is determined; the broadband azimuth spectrum function of each frame is determined according to the microphone signal, and the angle estimation value of each frame is determined;and the statistical result of each angle estimation value is determined according to the voice awakening confidence, and the angle estimation value of the maximum statistical result is the result ofDOA estimation. The angle estimation value of the maximum statistical result is the result of DOA estimation, and the statistical result of each angle estimation value refers to the statistical resultthat the statistical voice awakening confidence of each frame of microphone signal is greater than the threshold value, or the sum of the voice awakening confidence of the frame corresponding to thesame angle estimation value or the sum of the product of the voice awakening confidence of the frame corresponding to the same angle estimation value and the broadband azimuth spectrum function so that the azimuth angle of the voice signal source can be more accurately determined, and the signal-to-noise ratio can be enhanced and the voice damage can be reduced.

Description

technical field [0001] The invention relates to the field of voice processing, in particular to a method and device for improving DOA estimation of voice equipment. Background technique [0002] At present, smart voice hardware devices are more and more widely used, such as smart speakers and robots. These intelligent voice devices generally perform voice recognition after signal processing by a microphone array, thereby improving the voice recognition rate under far-field conditions. The functions that these devices generally have include being able to wake up with a certain keyword, find the direction of the speaker after waking up, and enhance the voice of the direction of the speaker, and can use it when playing music or voice. The wake word interrupts. Therefore, the main technologies involved include echo cancellation technology, direction of arrival estimation technology, beam forming technology, reverberation technology and so on. [0003] One of the problems exis...

AI Technical Summary

Problems solved by technology

[0003] At present, one of the problems existing in these intelligent voice interaction devices is the low far-field recognition rate.

The recognition of intelligent voice interaction equipment depends on the quality of the voice signal. The equipment has a relatively high recognition rate for the received clean voice signal, and a low recognition rate for the far-field voice signal affected by reverberation, noise, and interference.

This is because the current processing method is to perform DOA estimation when waking up. When multiple sound sources exist at the same time or when the device is far away from a strong reflective surface such as a wall or a display screen, the DOA estimation (based on the array-based sound wave arrival direction estimation) at the wake-up time is not effective. Accurate, causing the subsequent beamforming signal processing to mistakenly eliminate the speaker's voice as noise, resulting in the device being unable to understand the speaker's instructions

Images