A real-time talker localization implementation using MULTI-PHAT and particle filter

The estimation of a talker location in noisy and reverberant conditions using microphone arrays requires complex algorithms. However, the rapid increase in computational performance during the last two decades has opened the way for practical real-time talker localization applications. This paper presents an implementation of such system based on steered response power with phase transform (SRP-PHAT) combined with a particle filter (PF). The system utilizes the simplicity of SRP-PHAT and robustness of PF. The system is implemented using Pure Data software running on a standard laptop PC. Using realistic test data we show that accurate real-time talker location estimates can be produced even with a relatively low particle density.