Phillip L. Ainsleigh
Naval Res. Lab., Underwater Sound Reference Detachment, Orlando, FL 32856-8337
A wavelet-based method is introduced for estimating the arrival times of signals and echoes observed in noisy multipath environments. This problem arises when trying to model any signal containing overlapping segments of direct and reflected components. The method presented is more general than cepstrum-based techniques since it is not limited to echoes that are scaled replicas of the direct signal. The statistics of the wavelet coefficients are derived for pyramid-algorithm wavelet decompositions, and these are employed in a statistical detection algorithm for obtaining the arrival-time estimates. The method will be used to estimate signal and echo arrival times in data observed during underwater transducer calibration experiments conducted in small pressurized tanks.