A New Parametric Method for Time Delay Estimation
Maiza Bekara, Mirko van der Baan · 2007
In this paper, a new parametric method for time delay estimation is proposed. The method, classified under the generalized cross-correlation (GCC) approach, uses a couple of identical FIR filters to process the data from each sensor. A set of filters is designed to maximize the output cross-correlation at each lag. The lag associated with the maximum of all filtered cross-correlation is the estimate of the time delay, and its associated FIR filter is the optimum processor. The proposed method is implemented in the time domain and does not need spectral information as for the classical GCC methods. Its implementation uses eigen-decomposition algorithms and requires one input parameter which is the order of the FIR filter. It is equivalent to applying a data-driven bandpass filter to the cross-correlogram to emphasis the source signal. The proposed method is compared with standard methods in a simulation study. Simulation results show very good performance for the case of short data records and at low to moderate SNR levels.