A ConvLSTM-Based Blind Receiver for Physical Layer Wireless Communication

Recently, machine learning based blind detection of modulation and coding scheme (MCS) has been proven highly effective in achieving dynamic wireless communications in accordance with channel quality, thereby ensuring reliable and stable communication. To reduce the complexity of receivers using separate MCS detection and demodulation, we propose a ConvLSTM-based blind receiver to realize a unified solution with both blind MCS detection and low complexity demodulation. In achieving concurrent MCS detection and demodulation as a task of multi-label classification in deep learning, the ConvLSTM-based receiver includes three parts: data preprocessing, feature extraction, and multi-label classification. The data preprocessing is to guarantee a uniform input for all considered MCSs. Then, the feature extraction extracts the spatiotemporal features of the signals. Finally, the multi-label classification transforms the extracted features into bit streams. Simulation results show that the proposed ConvLSTM-based receiver can improve the bit error rate (BER) performance with low storage and time complexities compared with the baselines.