Sound source localization using SNR-based frequency weighting with microphone array

In order to improve the performance of sound source localization with microphone array in reverberant noisy environments, a frequency weighted sound source localization algorithm based on signal-to-noise ratio (SNR) is proposed. First, SNR of each frequency is estimated from the covariance matrix of the array signals in each frame. Then, the SRP-PHAT (steered response power-phase transform) formula is revised by the weight mapped from the frequency SNR by activation function. Finally, the SRP of each candidate location is calculated by the revised formula, and the sound source location is estimated by searching the maximum value of all the SRPs. The proposed method adaptively adjusts the contribution of frequency component to SRP according to the frequency SNR estimated in real time. Simulation results show that, compared with the conventional SRP-PHAT algorithm and Wiener pre-filtering beamformer algorithm, the proposed method obtains a higher percentage of correct estimates and a lower root mean square error, and is more robust against noise in complex noisy and reverberation environments.