Communication-aware Hankel singular value decomposition denoising method for underwater acoustic communication

This Letter proposes a communication-aware Hankel singular value decomposition denoising method (CA-HSVD) for underwater acoustic communication. The received passband signal is decomposed into singular components, whose communication relevance is evaluated using the normalized correlation-peak prominence of a known training sequence and mapped to continuous reconstruction weights through a sigmoid function. Semi-physical simulations using simulated communication signals, a measured sea-trial channel, and measured underwater acoustic noise show that CA-HSVD achieves the lowest bit error rate (BER) over the tested signal-to-noise ratio (SNR) range and provides a SNR gain of approximately 2 dB over training-guided hard-Hankel singular value decomposition at the same BER level.