Improving multipath time delay estimation in shallow water with power spectrum weighted auto-correlation

It's difficult for the cepstrum method to extract the multipath component of an underwater sound source at short range since the component becomes too weak. To deal with this problem, a power spectrum weighted auto- correlation multipath time delay estimation and its automatic extraction method is proposed. This method is carried out as follows:Firstly, the power spectrum of input signal is estimated with the FFT method; Secondly, the power spectrum is powered with 1-α (0 〈 α 〈 1); Thirdly, the IFFT is performed to transform the above data back to time domain. Lastly, a peak detection processing is performed to obtain the multipath time delay. When the preceding time delay is determined, the following time delay is searched within an interval around the preceding one. Theoretical analysis show that, the cepstrum and auto-correlation methods of multipath time delay estimation are two specifications of the power spectrum weighted auto-correlation method. By sea trial data processing and simulating with Bellhop ray model, the proposed method is proved to obtain stronger multipath trace than cepstrum method. Simulation also shows that the proposed method improves the output SNR of the multipath component by 3~4 dB. The bottom path time delay extracted with the proposed method is used in source range estimation with only a single hydrophone. The result is conformed basically to the range measured with GPS, which validated the effectiveness of our method.