Target Direction Speech Detection Method Based on Second-Order Cone Programming

Abstract

The invention relates to a target direction voice detection method based on second-order cone programming, comprising the following steps: designing the beamforming of the lowest sidelobe for the target direction, and calculating the weight of the beamforming of the lowest sidelobe by using the second-order cone programming; constructing Noise estimation weight; Estimate the target signal and noise signal; Calculate the power of the target signal and the noise signal by using the first-order recursive smoothing in the time domain; Use the power of the target signal and the power of the noise signal to calculate the posterior signal-to-noise ratio; The minimum value tracking of the posteriori signal-to-noise ratio is carried out to obtain the minimum value of the posteriori signal-to-noise ratio; the sum of the posteriori signal-to-noise ratio and the minimum value of the posteriori signal-to-noise ratio in the frequency range of 281.25Hz to 3437.5Hz are calculated and the ratio of the sum; judging the magnitude of the ratio and the set threshold to determine whether the voice in the target direction exists. The detection method of the present invention has the advantages of fewer steps and less calculation amount, and the problem of instability of some frequency points can be avoided through frequency domain summation.

Description

technical field [0001] The invention relates to the technical field of target direction voice detection, in particular to a target direction voice detection method based on second-order cone programming. Background technique [0002] Speech detection in the target direction is a technology that can determine whether the voice in the target direction exists at the current time. It plays an important role in human-computer interaction, speech enhancement, and far-field speech recognition. [0003] The traditional method commonly used statistical model VAD (Voice Activity Detection, Voice Activity Detector), this method can distinguish whether there is voice at the current time, but the required constraints are single sound source, stable noise, high signal-to-noise ratio It works in some cases, and there is no way to use spatial information to determine which direction the current voice is coming from. Moreover, in the far-field environment, the effect of VAD will be greatly ...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to overcome the defects of the prior art, to provide a target direction voice detection method based on second-order cone planning, to solve the problem that the VAD in the traditional method only distinguishes whether there is voice at the current time and lacks spatial judgment, and in the far field environment It also solves the problem of inaccurate target direction estimation and easy failure in the case of unstable noise in the heuristic SNR-based method.

Images