Reliability Measurement Of Time Difference Of Arrival Estimations For Multiple Sound Source Localization

Time Difference Of Arrival (TDOA) estimates are used for passive acoustic single sound source localization with microphone arrays. The technique of choice in most systems for TDOA estimation is the Generalized Cross-Correlation (GCC) method [1]. For a multi-sound source scenario, cross-correlation terms of the active sound source signals as well as noise and reverberation effects complicate the TDOA estimation and make a robust sound source localization difficult. In this publication a method increasing the absolute correct TDOA estimate percentage by up to 5.06% compared to the standard method is presented. Furthermore, we demonstrate that two properties of the GCC function - the value of the maximum peak and the ratio between the the##### peak - can be used as reliability indicators for the TDOA estimates. Separating these confidence criteria into intervals with increasing values, a highly confident correct TDOA estimate percentage of about 98% in the interval with highest values could be achieved for both criteria. This study is carried out with real data recorded for a stationary two sound source scenario in a noisy and reverberant office room environment