Experimental evaluation of the joint position-pitch estimation (POPI) algorithm in noisy environments

This paper proposes an enhancement and evaluates the performance of a recently proposed method for joint extraction of pitch and direction of arrival for speaker localization. We propose a new weighting function in order to suppress the cross-terms of the aforementioned method. The performance of the method is evaluated by measuring the correct estimate of position at a frame level. This evaluation analyzes the limits of the method at different background noise levels by using real world recordings. The results show that the proposed algorithm gives more consistent location estimates leading to a reduction in angular deviation by 4.23deg at -3 dB SNR.