Classification of rotor blade number of rotor targets micro-motion signal based on CNN

In this paper, convolutional neural network (CNN) is used to classificate the rotor blade number of rotor targets micro-motion signal with deep learning’s strong feature extraction ability. Firstly, the scattering point model of the rotor blade echo is used to generate the target echo. Under the condition of different signal-to-noise ratio, time-frequency diagram of the echo with different number of rotor blades is constructed by using short-time Fourier transform, which is used as the test set and training set. Three convolutional neural network models of lenet, alexnet and vggnet are used for training. The performance of the network model is compared, and the recognition performance of the alexnet network model is analyzed under ambiguous, unambiguous and a method of Interpolation to resolve ambiguous. Through experiments, it can be found that the recognition rate of the proposed method can reach 95% under the condition of signal-to-noise ratio of 10dB. It has good recognition performance for classification of rotor blade number, and provides effective data and algorithm support for the rotor target recognition in the future.