Voice enhancing algorithm based on second-order differential microphone array

Abstract

The invention provides a voice enhancing algorithm based on a second-order differential microphone array, and belongs to the field of voice signal processing. The method comprises the following steps:first, establishing the microphone array, collecting three paths of voice signals of voice of a speaker, extracting target voice beam forming signals and noise beam forming signals by adopting a second-order differential algorithm, carrying out framing and frequency band dividing, randomly selecting one path of voice signals, carrying out framing and frequency band dividing, calculating the masking value of each time-frequency unit, and carrying out smoothing processing, thus obtaining the time-frequency unit value of voice after enhancement of each time-frequency unit; and finally, through inverse Fourier transform, carrying out windowing and overlap-adding, thus obtaining the enhancing signal corresponding to the voice of the speaker. According to the method, by combining with the beamforming algorithm and a computational auditory scene analysis algorithm, the beam forming result is only adopted for estimating the target voice and noise energy, the masking value generating processin the computational auditory scene analysis is optimized, so that the masking vale is more smooth and is suitable for a practical application scene, and finally, the enhancing effect after voice synthesis is obvious.

Description

technical field [0001] The invention belongs to the field of digital voice signal processing, and in particular relates to a voice enhancement algorithm based on a second-order differential microphone array. Background technique [0002] With the development of electronic information technology, speech signal processing technology has been widely used in speech interactive systems. However, due to various noises in the daily environment, the quality of speech processing is greatly affected, which is mainly reflected in the decline of speech recognition rate and intelligibility. Therefore, in speech signal processing, it is necessary to perform preprocessing through speech enhancement technology to remove noise and interference in speech, thereby improving the quality of speech processing. [0003] The speech enhancement algorithm based on the second-order differential microphone array is more conducive to achieving a speech noise reduction effect with good directivity. Amo...

AI Technical Summary

Problems solved by technology

[0024] 1. The beamforming algorithm can effectively use the time difference of signals collected by different microphones for high-directional denoising, but the result obtained in the formula is different from the frequency term ω 2 Correlation, if the correlation filtering operation is not performed, the output signal will be severely distorted as the frequency changes, and the design of the correlation filter is more complicated and difficult to achieve in hardware implementation

[0025] 2. Calculation The IRM calculation method in the auditory scene analysis algorithm is relatively rigid. First, the estimation of the target voice and noise energy may not be accurate, which may affect the estimation of the masking value. Secondly, in practical applications, there is often a large gap between the target voice and the direction of the noise source. , and the time distribution is not uniform, it is difficult to achieve optimal performance by using the IRM value directly as a masking value for weighting on all time-frequency units

Images