Deep Neural Network Based Parallel Signal Detection in SM-OFDM System

초록

A novel deep neural network based parallel signal detection (DNN-PSD) is proposed for the spatial modulation based orthogonal frequency division multiplexing (SM-OFDM) system. With the purpose to reduce the complexity of the conventional DNN, a uniform small-scale DNN with fewer parameters and less training time is exploited to detect the signals for each subcarrier parallelly. Apart from maximum likelihood (ML) and maximal ratio combining (MRC) detection schemes, the detailed DNN-PSD algorithm and its complexity analysis are presented. Simulation results confirm that the bit error rate (BER) performance of the proposed DNN-PSD is far superior to the MRC detection and similar to the optimal ML detection but with much lower complexity under different scenarios. It has more robustness and achieves a finer compromise between BER performance and complexity.

키워드