Deep Convolutional Neural Network-Based Detector for Index Modulation
Tengjiao Wang, Fang Yang, Jian Song, Zhu Han
Abstract
In this letter, a convolutional neural network-based detection framework is proposed for the wireless communication systems using orthogonal frequency-division multiplexing with index modulation (OFDM-IM). In the proposed framework, the received symbols are transformed to polar coordinates to help the neural network detect the indices of the activated subcarriers. We parallel the amplitude and the phase of the received symbols to form 2-dimensional matrices and use 2-dimensional convolutional layers to fully exploit the inherent information in the OFDM-IM symbols. After offline training, the proposed detector can be employed to implement online detection of the OFDM-IM symbols. Simulation results demonstrate that the proposed detector is capable of achieving near maximum likelihood detection performance with much lower complexity.