Fingerprint Identification of Short Wave Transmitter Based on Deep Learning

摘要

Traditional transmitter identification method needs manual construction of fingerprint features, and it is difficult to get good results in the actual battlefield environment. Aimed at this problem, this paper proposes a fingerprint identification method for shortwave transmitter based on deep learning. The method extracts the signal fingerprint feature by convolution and pooling operation, and combines with Softmax classifier to realize the fingerprint feature recognition. Firstly, the signal data of the transmitter are collected, and divided into training data and test data, then the data are transformed by the rectangular integral bispectrum, and bispectrum feature is taken as the feature input vector of network training. Secondly, establish a convolutional neural network model, determine the network structure, parameters of each layer, learning rate and so on; Finally, input training data to train the network, use backpropagation to adjust network parameters and verify network performance with test data. The experimental results show that this method improves the accuracy of fingerprint feature recognition effectively, and the recognition performance is better than the traditional method.