COMPARISON OF DISCRETE SUBSAMPLE TIME-DELAY ESTIMATION METHODS APPLIED TO NARROW-BAND SIGNALS

The problem of estimating the subsample time delay between two noise-c orrupted narrowband signals is considered. Significant improvements to two previously proposed estimators, the MXS and MSX, are made. The pe rformance of the revised estimators is assessed through simulation ove r a range from -10 to 20 dB, with four well known estimators, two of w hich are asymptotically maximum likelihood (the direct correlator and the average square difference function) and the other two are windowed versions of them (the generalized cross-correlator and generalized le ast squares). Also, the estimators are applied to ultrasonic experimen tal data at three different SNRs. At low SNR, the MXS/MSX estimators a re shown to exhibit a significant reduction in the variance of the est imates.