Deep Learning based Underwater Acoustic OFDM Receiver with Joint Channel Estimation and Signal Detection

Abstract

This paper presents a deep learning (DL) based underwater acoustic (UWA) orthogonal frequency division multiplexing (OFDM) receiver, which exploits both DL and expert knowledge of block-based signal processing methods in wireless communication to improve system performance and interpretability. The proposed receiver jointly employs skip connection (SC) convolutional neural network (CNN) for channel estimation and bi-directional long short-term memory (BiLSTM) network for signal detection, abbreviated as SCBNet. Specifically, the channel estimation subnet is designed with SC CNN to utilize the thought of image super-resolution to reconstruct the entire channel frequency response of all subcarriers. The signal detection subnet is designed with BiLSTM to extract the correlations of received sequential data for signal detection. The proposed SCBNet was evaluated by experimental data, and the results have demonstrated that the SCBNet has the lowest BER performance compared with traditional linear algorithms, DL-based black-box receiver, and ComNet receiver.