Learning-Based Signal Detection for OFDM Systems with I/Q Imbalance

Abstract

The in-phase/quadrature (I/Q) branches imbalance leads to mirror subcarrier interference and worsens the performance of zero-intermediate frequency (zero-IF)-based orthogonal frequency division multiplexing (OFDM) systems. To tackle the signal detection issue of in-phase/quadrature imbalance (IQI) at the transceiver, a deep learning-based approach is proposed by using the convolutional neural network. Specifically, the network model and parameters are well-designed based on the features of the channel impulse response and the mirror interference of IQI. To verify the designed model, it is first trained by simulated data under the off-line training and then used directly to recover the on-line transmitted data. The simulation results demonstrate that the proposed method shows excellent performance in processing OFDM signals under the case of IQI and it is more robust than traditional methods even without cyclic prefix.