Signal Recognition Method of X-ray Pulsar Based on CNN and Attention Module CBAM

Fast and accurate identification of X-ray pulsar signals is an important prerequisite for pulsar navigation. Most of the current identification methods are to extract the cumulative profile features and then compare them with features of the standard profile to complete the identification. However, the profile with high signal-to-noise ratio needs long observation time, which has a great impact on real-time identification. In this paper, the X-ray pulsar signal is converted into time interval sequences, and then the feature extraction and identification are completed by using one-dimensional convolution neural networks. In terms of network architecture design, we introduce convolutional block attention module (CBAM) and propose the CBAM-Inception module to construct the network. This structure combines the advantages of Inception and CBAM, and uses channel and spatial attention mechanisms to enhance the feature extraction capabilities of the Inception network. Experimental shows that the proposed method can greatly shorten the required observation time while ensuring high-accuracy X-ray pulsar signal identification. Moreover, the comparison of convolution blocks shows that the CBAM-Inception block can greatly improve network identification ability.