Experimental comparison of cross correlation-based time-delay algorithms for acoustic source localization in real environment

Time delay estimation (TDE) is one of the primary techniques for speaker localization based on microphone array in many applications. The cross-correlation method is one of the basic solutions of TDE problems. The weighted generalized cross-correlation (GCC) based on this theory and relied on the spectral characteristics of the signal, is a very widely used algorithm. In the real environment, background noise and channel multipath due to room reverberations are the two major factors which cause algorithms degradation. In this paper, we present another method which is also based on cross-correlation. It uses the vocal glottal excitation information, which is an essential characteristic of voiced speech and be very robust in the complex environment. Previous simulations have shown the two methods are both applicable. In this paper, these two algorithms are tested in real ordinary laboratory. The experiments focus on the comparison of the accuracy affected by the observation interval, the spatial structure of microphone array between the two algorithms. Extensive experiment results reveal that the expected glottal excitation information based method performs better than the GCC method in real environment.