Accurate Multicarrier Waveform Classification Using Convolutional Neural Networks

To improve spectrum utilization efficiency and flexibility in diverse application scenarios, several new multicarrier modulation techniques have been proposed in recent years, and thus accurate recognition of various multicarrier waveforms is critical in heterogeneous wireless networks. After analyzing their different characteristics of popular multicarrier waveforms, this paper designs a new classification scheme based on 4-layer convolutional neural networks. In particular, the Fourier synchrosqueezing transform and Haar wavelet transform are exploited to discriminate different multicarrier waveforms. The proposed approach does not require any priori knowledge of the received signals, unlike the traditional methods. Simulation results demonstrate that the proposed classification scheme outperforms the benchmark schemes in the literature, in terms of the classification accuracy even at low signal-to-noise ratio.