Confidence Scoring of Time Difference of Arrival Estimation for Speaker Localization with Microphone Arrays

Microphone arrays can be employed for passive acoustic source localization using time difference of arrival (TDOA) estimates in microphone pairs. The most common method for this TDOA estimation is the generalized crosscorrelation (GCC) method which is also used in this work. In an office room environmental noise influences and reverberation effects complicate the TDOA estimation and aggravate a robust sound source localization. This publication presents for a stationary single source scenario with real data a method which increases the correct TDOA estimate percentage to 70.30% compared with 61.80% for the standard method of estimating the TDOAs. Furthermore, we show that two properties of the GCC function can be used as confidence criteria of the TDOA estimates: the value of the maximum peak as well as the ratio between the and the peak in the GCC function are appropriate indicators for the reliability of a TDOA estimate. Separating these confidence criteria into intervals with increasing values shows for the interval with highest values a very reliable correct TDOA estimate percentage of about 95% for both criteria.