A PROBABILISTIC FRAMEWORK FORMULTIPLE SPEAKER LOCALIZATION

This paper presents a novel probabilistic framework for local-izing multiple speakers with a microphone array. In this frame-work, the generalized cross correlation function (GCC) of each mi-crophone pair is interpreted as a probability distribution of the time difference of arrival (TDOA) and subsequently approximated as a Gaussian mixture. The distribution parameters are estimated with a weighted expectation maximization algorithm. Then, the joint dis-tribution of the TDOA Gaussian mixtures is mapped to a multimodal distribution in the location space, where each mode represents a po-tential source location. The approach taken here performs the lo-calization by 1) reducing the search space to some regions that are likely to contain a source and then 2) extracting the actual speaker lo-cations with a numerical optimization algorithm. The effectiveness of the proposed approach is shown using the AV16.3 corpus. Index Terms — Microphone arrays, localization, multiple speakers, Gaussian mixture, steered response power