A method for personal identification of communication radiation source based on deep belief network

Aiming at the problem of individual identification of communication radiation sources in complex electromagnetic environment, we propose a method of mutual modulation interference recognition of communication radiation sources based on deep confidence network under small sample conditions. Firstly, we analyze the amplitude and phase characteristics of intermodulation interference of communication radiation sources, which can be used as individual characteristics to distinguish communication radiation sources. Then, the square integrated bispectra of communication radiation source adopts contrast divergence method to train each restricted Boltzmann machine from the bottom up, through which the appropriate weights, the deviation of the hidden layer and the deviation of the visible layer are obtained, which represent the intermodulation interference characteristics of the radiation source signal.Finally, the training model is fine-tuned by softmax classifier to obtain a deep learning network for the fine feature recognition of communication radiation sources.The recognition rate of more than 80% is obtained by computer simulation, which further validates the effectiveness of the method.