Optimization of Deep Learning Technique for OFDM Receivers in 6G Wireless Communications

INTRODUCTION: This paper presents an innovative deep learning-based optimization technique for orthogonal frequency division multiplexing (OFDM) receivers in wireless communication systems. OBJECTIVES: The proposed method utilizes an enhanced deep convolutional neural network (Enhanced DCNN) architecture with a time-frequency domain fusion mechanism to address the issues of interference and temporal information loss. The model incorporates attention mechanisms and causal convolutions to extract long-term dependencies within the received OFDM signals. It enables accurate channel estimation and signal recovery. METHODS: The methodology is validated using simulations based on 3GPP-defined channel models. It includes extended typical U (ETU), extended pedestrian A (EPA) and extended vehicular A (EVA) across varying signal-to-noise ratio (SNR) conditions. RESULTS: Results demonstrate that the proposed receiver significantly improves bit error rate (BER) performance compared to traditional Least Squares (LS) and LMMSE methods. Particularly, in scenarios with large delay spreads and high mobility. Additionally, the model has a lower computational complexity (CC) and thus is appropriate for real-time implementation. CONCLUSION: We view this work as a strong scheme to improve the performance of OFDM systems in future wireless networks.