Height approximation for audio source localisation and tracking

The stochastic region contraction (SRC) algorithm has been proposed in the literature as a method for acoustic localisation using a microphone array in a noisy and reverberant environment. This technique makes use of the steered response power (SRP), a costly but robust technique for source localisation, and finds the global maximum vastly more efficiently than using a grid search method. We discuss combining this technique with prior information (e.g. in future work we will use a video tracker) to speed up the algorithm by, in some cases, an order of magnitude by limiting the heights to be searched. This gain is derived from simulations and is achieved whilst at the same time not neglecting large search volumes, continuing to allow a change of audio sources to be detected.