Complex-valued Deep Network aided Channel Tracking for Underwater Acoustic communications

In this work, a complex-valued deep neural network (ℂ-DNN) aided channel tracking approach is proposed for underwater acoustic (UWA) orthogonal frequency division multiplexing (OFDM) communications. The experiments with the at-sea-measured WATERMARK dataset indicate that, the complex-valued model can perform a near-optimal channel tracking, as well as saving 50% spatial resources compared to the real-valued counterpart. Besides, the model transfer with UWA environment mismatch is also validated in this paper.