Time-frequency Overlapped Signals Intelligent Modulation Recognition in Underlay CRN

In underlay cognitive radio networks (CRN), a novel intelligent modulation recognition method of time-frequency overlapped MPSK/MQAM signals based on convolutional neural network (CNN) is proposed in this paper. Aiming to achieve better modulation recognition performance of the time-frequency overlapped signals, the contour maps of cyclic spectrum of the received signals as the initial features are extracted, and then the CNN is used to extract further features as the final features. After preprocessing the contour maps to serve as input data for the CNN, the CNN is constructed and optimized to complete the training of the CNN. Finally, the well trained CNN is used for modulation recognition of the time-frequency overlapped MPSK/MQAM signals. Simulation results show that the average recognition rate of the proposed method is 90 % when the signal-to-noise ratio (SNR) is 2dB, and it is robust to the power ratio and spectrum overlapped rate of the component signals.