Learn to Detect and Detect to Learn: Structure Learning and Decision Feedback for MIMO-OFDM Receive Processing

One major open challenge in MIMO-OFDM receive processing is how to efficiently and effectively utilize the extremely limited over-the-air pilot symbols to detect the transmitted data symbols. Recent advances have been devoted to investigating effective ways to utilize the limited pilots. However, we notice that besides exploiting the pilots, one can take advantage of the data symbols to improve detection performance. Thus, this paper introduces an online subframe-based approach, namely RC-StructNet-DF, that can efficiently learn from the precious pilot symbols and be dynamically updated with the detected payload data using the decision feedback (DF) approach. With the DF mechanism, the network can dynamically track the channel changes within a subframe. To mitigate the error propagation of the DF approach, the specially designed StructNet is adopted in the frequency domain, which is shown to be robust to the incorrect labels owing to the embedded structural information. The introduced parameter estimation (PE) layer in the StructNet further facilitates the DF method by utilizing the network weights to learn the parameters. Extensive experiments have been conducted to demonstrate the effectiveness of RC-StructNet-DF in detection in both the MIMO-OFDM system and the massive MIMO-OFDM system with different modulation orders under various scenarios.