Mohammad K. Nehal
Juan A. Henriquez
Terry E. Riemer
Russell E. Trahan
Dept. of Elec. Eng., Univ. of New Orleans, New Orleans, LA 70148
Uniform and multiple delays/advances were estimated under heavy noisy conditions [signal-to-noise-ratio (SNR(corresponds to)signal energy/noise energy) below 0 dB]. The technique introduced, called the spatial amplitude mapping (SAM) method, isolates a data segment from each of the channels in a multichannel system by using a suitable window. When two matched segments are plotted on an x--y plane, the distribution of every pair of windowed coordinates will remain near a straight 45(degrees) line that passes through the origin. The distribution of two nonmatched segments or two matched segments containing noise will scatter around this line; hence, a pair of matching windowed segments can be found by searching for the distribution closest to the 45(degrees) line. This information is then used to estimate the segment delays. Under very noisy conditions, however, multiple delays can be detected. A recursive least-squares (RLS) filter is then used to adaptively estimate the correct delay. The technique was implemented in the delay estimation of synthetic (stationary) data and neurophysiological (nonstationary) data with satisfying results. Compared to the window correlation technique (WCT) [Callison et al., J. Acoust. Soc. Am. 81, 1000--1006 (1987)], SAM can estimate delays down to a SNR of -17.1 dB while the lower bound for the WCT is -2 dB.