Noisy input-output system identification approach for time delay estimation

An unbiased impulse response estimation approach is applied to time delay estimation between noisy signals received at two spatially separated sensors. The estimated delay is modeled by an adaptive finite impulse response filter whose coefficients are updated according to a modified least mean square (LMS) scheme. Simulation results show that the proposed algorithm outperforms the conventional LMS time delay estimator for autoregressive signals as well as real speech data. (C) 2002 Elsevier Science B.V. All rights reserved.