Array sound source positioning method based on signal-to-noise ratios of frequency points and offset soft judgment

Abstract

The invention discloses an array sound source positioning method based on signal-to-noise ratios of frequency points and offset soft judgment. A sound source signal is collected through a microphone array; the sound source signal is divided into single-frame signals in a time domain; the single-frame signals are converted into frequency domain signals through windowing and discrete Fourier transform; the single-frame signals are continuously segmented into sub-frame signals; the sub-frame signals are converted into frequency domain signals through the windowing and discrete Fourier transform;the signal-to-noise ratios of the frequency points are calculated and mapped into offset soft judgment values; corrected controllable response power values of candidate positions are calculated; and the candidate position with the maximum controllable response power value is estimated as a sound source position. The array is composed of microphones, so that three-dimensional positioning is realized; the signal-to-noise ratio is estimated for each frequency point of the whole frequency band; direct positioning is carried out without considering a time delay difference between the time of reaching the two microphones by a sound source, and a Sigmoid function is used for increasing an offset to realize soft correction; the specific frequency points are restrained by reference to the signal-to-noise ratios; prior information of the signal statistical property is not needed, and a noise power spectrum does not need to be estimated by using data of mute frames; the method is suitable for anenvironment with relatively quick noise transformation; the indexes including a positioning success rate and a root-mean-square error (RMSE) are better; and the positioning success rate is obviously improved in medium and low-signal-to-noise ratio environments, and the algorithm robustness is better.

Description

technical field [0001] The invention relates to the technical field of signal processing, in particular to a sound source localization method. Background technique [0002] The controllable response power SRP-PHAT sound source localization algorithm weighted by phase transformation does not require prior knowledge of signal and noise, and only needs 20-30ms data to estimate the position of the sound source. It is a sound source localization that can be realized in real time The algorithm has strong anti-reverberation ability, but the traditional SRP-PHAT sound source localization method is not satisfactory in the complex acoustic environment, especially when the noise increases, the localization performance drops significantly. [0003] In order to improve the success rate of sound source location in noise and reverberation environment, this method adopts frequency point SNR and offset soft decision, estimates the signal and noise power spectral density of each frequency poi...

AI Technical Summary

Problems solved by technology

[0002] The controllable response power SRP-PHAT sound source localization algorithm weighted by phase transformation does not require prior knowledge of signal and noise, and only needs 20-30ms data to estimate the position of the sound source. It is a sound source localization that can be realized in real time The algorithm has strong anti-reverberation ability, but the traditional SRP-PHAT sound source localization method is not satisfactory in the complex acoustic environment, especially when the noise increases, the localization performance drops significantly

Images