Blind deconvolution of extended duration underwater signals

Abstract

Synthetic time reversal (STR) is a passive blind deconvolution technique for estimating the original source signal and the source-to-array impulse responses in an unknown multipath sound channel. Previous investigations of STR involved relatively short duration chirp signals (50 ms) where the assumption of a static underwater sound channel was acceptable. However, the static-channel assumption is inadequate for longer duration underwater acoustic communication signals lasting one or more seconds. Here, wave-driven changes in the ocean's surface shape and water-column sound speed lead to significant temporal variations in the source-to-array impulse responses. This presentation describes an effort to accommodate such temporal variations by decomposing a long duration signal into smaller overlapping pieces, applying STR to each piece, and then stitching the resulting sequence of signal estimates together to blindly recover the original long-duration signal. Simulations of a one-second-duration synthetic signal propagating in a static underwater sound channel to a 16-element vertical array are used to determine how the technique's performance depends on the duration and overlap of the signal pieces in relation to the signal's bit rate and the sound channel's time-delay spread. Extension of this effort to recordings from dynamic sound channels is anticipated. [Sponsored by the Office of Naval Research.]