Joint Identification and Localization of a Speaker in Adverse Conditions Using a Microphone Array

We discuss a joint identification and localization microphone array system based on diagonal unloading (DU) beamforming, which has been recently introduced for acoustic source localization. First, we propose a DU beamformer version for the signal enhancement problem. Then, we propose a enhanced DU steered response power (SRP), in which the first estimate of the source position is further refined with the information gathered from the speaker recognition module. The enhanced SRP-DU is obtained by weighting the frequency components with respect to the spectral characteristics of the speaker. The approach does not add significant computational load to the array processing. Experiments conducted in noisy and reverberant conditions show that the use of the DU beamformer provides better speaker recognition performance if compared to the conventional one since it reduces deleterious effects due to the spatially white noise and point-source interferences. Simulations also show that the speaker identification can improve the localization accuracy, and it is thus interesting for applications and systems which rely on integrated localization and speaker identification.