Spectrum Sensing Based on Spectrogram-Aware CNN for Cognitive Radio Network
Lianning Cai, Kaitian Cao, Yongpeng Wu, Yuan Zhou
Abstract
Spectrum sensing is one of the key problems in the cognitive radio network. Existing spectrum sensing methods commonly use deep learning models such as the convolutional neural network (CNN) and the long short-term memory network (LSTM). In this letter, we take the spectrogram of signal samples obtained by short-time Fourier transform as the input of CNN and propose a spectrogram-aware CNN (S-CNN) algorithm. In addition, to further improve the generalization of the CNN model, we adopt the data augmentation technique based on a deep convolutional generative adversarial network to generate additional training data. Simulation results show that the proposed S-CNN algorithm outperforms the CNN and LSTM-based methods in terms of detection performance.