Electromagnetic signal identification method based on depth long-short-term memory network

The invention discloses an electromagnetic signal identification method based on a depth long-short-term memory network, and the electromagnetic signal identification method comprises the steps: carrying out the preprocessing of outlier removal, sliding window slicing and normalization of electromagnetic signal sequence data, and enabling the electromagnetic signal sequence data to serve as an input feature vector; setting a depth long-short-term memory network model according to the input feature vector of the electromagnetic signal, the first layer being an input layer, the middle layer being a long-short-term memory layer, and the last layer being a full-connection output layer, and using a normalization exponential function as final output; performing iterative optimization on the lossfunction through a back propagation optimization algorithm, and updating the network weight coefficient towards the reduction direction of the loss function to obtain a network model for electromagnetic signal identification