Deep Bidirectional LSTM Network Learning-Aided OFDMA Downlink and SC-FDMA Uplink

In this paper, deep learning (DL)-aided signal detection is proposed for the orthogonal frequency division multiple access (OFDMA) in the downlink and for the single carrier frequency division multiple access (SC-FDMA) in the uplink. A deep bidirectional long short-term memory (D-BiLSTM) is considered for training OFDMA/SC-FDMA symbols. The traditional orthogonal frequency division multiplexing (OFDM) receiver explicitly estimates the channel state information (CSI) and recovers the transmitted symbol with the aid of the estimated CSI. By contrast, the proposed scheme implicitly estimates the CSI, and the receiver directly detects the transmitted symbols. Our simulations results show that the DL-based system is capable of addressing the channel distortion and can recover the transmitted data. A DL model is trained offline in order to mitigate channel distortion. Following the training, the model is tested to recover the transmitted online data symbol directly. Our simulations demonstrate that BiLSTM based DL may be effectively employed to detect signals in the OFDMA downlink and the SC-FDMA uplink scenarios.