Adaptive recognition of modulated signals based on complex convolutional neural networks

Adaptive recognition and of modulated signals play an important position in modern wireless communications. In order to solve the problem of low recognition accuracy of existing models, we propose an complex convolution neural networks(SACNN). Firstly, the modulated signal data is preprocessed using a short-time fast Fourier transform. Then, a self attention mechanism layer is added after the CNN convolutional layer, which utilizes the similarity between features to increase the weight of corresponding features, thereby achieving the goal of improving the accuracy of target recognition. The final experimental results show that the model is able to accurately recognize 11 different modulated signals at high signal-to-noise ratio, and the average recognition rate can reach up to 90.5% without confusion between different modulated signals, and has higher robustness and higher accuracy than other deep neural networks.