A note on the modified and mean-based steered-response power functionals for source localization in noisy and reverberant environments
Máximo Cobos · 2014
One of the most important applications of microphone arrays is the localization of one or multiple speakers in reverberant and/or noisy scenarios. The steered-response power (SRP) method, which searches for the maximum accumulated power by following a grid search procedure, is one of the leading algorithms for performing this task. However, the computational requirements of this algorithm are strongly increased when using many microphones or very fine spatial grids. The modified SRP (M-SRP) was proposed to tackle this problem, improving the performance of the original method when coarser spatial resolutions are considered. Recently, a mean-based variant of the M-SRP (MM-SRP) and an iterative search procedure were also introduced, allowing for higher accuracy with an additional reduction in the number of functional evaluations. This paper discusses further the differences among all these functionals, providing deeper insight into their motivation and performance in adverse acoustic environments.