INTERSPEECH 2012
13th Annual Conference of the International Speech Communication Association

Portland, OR, USA
September 9-13, 2012

TDOA Estimation for Multiple Speakers in Underdetermined Case

Mariem Bouafif (1), Zied Lachiri (1,2)

(1) LSTS-SIFI Laboratory, National Engineering School of Tunis, Tunis, Tunisia
(2) Depart. of Physics and Instrumentation, National Institute of Applied Sciences and Technology, Tunis, Tunisia

In this paper we address the issue of estimating the time delay of arrival in underdetermined case. We develop a method using the excitation characteristics of the speech production. This method is based on the cross correlation of the Hilbert Envelops of linear prediction residuals derived from two microphones signals. The method has been applied to real data obtained by recording many sources captured by a pair of microphones. Experiments show that reverberation distorts the input signals, each reverberation causes an extra peak in the crosscorrelation. This makes it difficult to determine which peak is the central time-delay peak and which are just reverberation sidelobes. An alternative time delay estimation method has been implemented and compared to spectrum angular methods.

Index Terms: TDOA, Linear prediction, Hilbert Envelope

Full Paper

Bibliographic reference.  Bouafif, Mariem / Lachiri, Zied (2012): "TDOA estimation for multiple speakers in underdetermined case", In INTERSPEECH-2012, 1748-1751.