Robust Speaker Localization Utilizing a Novel Beamforming Algorithm Based on Harmonic Structures

Speaker localization by microphone array has recently received significant attention. Although various methods have been proposed; their performance with short data segments under noise and reverberation degrades considerably. Sound localization based on Steered Response Power (SRP) shows more robustness in practical situations especially with the use of short data segments. In SRP-PHAT algorithm by employing a filtering scheme based on phase transform, performance of this localization algorithm is improved in reverberant environments. There has been relatively little success at applying an SRP filtering scheme that is robust in both noisy and reverberant conditions. This paper presents a novel and useful filter for SRP localization. The proposed approach employs estimated periodicity of harmonic spectral intervals of speech signal to design a filter for each channel of microphone array and is used within the filter and sum beamforming algorithm. Simulation results in various acoustical conditions (different SNR values and reverberation parameters) shows the superiority and robustness of the new algorithm compared to the conventional methods