Sparse acoustic response function estimation via Gaussian‐mixture model with application to M‐ary spread spectrum communications.

Abstract

An important feature of underwater acoustic response functions is their sparsity in delay‐Doppler. Models of the response must have the complexity to exploit this sparsity and the flexibility to adapt to the diverse environmental conditions that impart it. The Gaussian‐mixture assignment over delay‐Doppler slots can provide such a framework and leads to adaptive shrinkage operators that effectively attenuate small noise‐like delay‐Doppler paths and leaves larger occupied paths unaffected [Gendron, IEEE Trans. SP 53 2005]. Accurate treatment of the latent parameters of the model aids performance across environments and extends its usefulness in underwater acoustic applications. This model is applied to M‐ary orthogonal spread spectrum signaling where necessary SNR/bit at low received SNR is obtained with array, processing, and coding gain. The approach allows for bulk path dilation estimation and compensation that extends the effective channel coherence duration to that associated with the natural Doppler variations between the various acoustic paths. Symbol decisions are iterative with channel estimates and bulk time varying dilation compensation. These receiver structures are tested on very shallow water acoustic transmissions in Buzzard’s Bay MA. Probability of bit error is reported under 10−5 with two elements combining at −16 dB received SNR.