Wideband processing of acoustic signals using wavelet transforms. part II. Efficient implementation and examples

In Part I [J. Acoust. Soc. Am. 96, 850-856 (1994)], a theoretical derivation for processing wideband signals to localize sources was presented. This was first done for the active case of a single sensor illuminating the environment. It was then done for the passive case of three receivers sensing the environment. In both cases, the processing was formulated as a deconvolution. This part presents a new and efficient technique for coherently processing signals received at three or more passive sensors, and therefore implementing the wideband passive deconvolution to estimate the wideband spreading function. The technique is first described for the case of three receivers passively sensing an environment; it is then generalized to N passive sensors. The issue of additive noise on the received signals is then addressed. The noisy received signals are processed with the new implementation of the deconvolution. An expression relating the signal-to-noise ratio (SNR) of the estimate of the wideband spreading function to the SNRs of the received signals is also derived. The results show that the wideband deconvolution robustly operates on signals that have been contaminated with noise. Examples are provided.