Weaving missile classification using deep learning

In target tracking estimation of maneuvering targets, the estimation of an unknown weaving target frequency is crucial for improving the miss distance. In this paper, we propose estimating the weaving frequency using deep neural networks, instead of classical estimation algorithms based on the Kalman filter framework. After designing a network, we further compare the proposed approach performance with multiple model adaptive estimation. Simulation results show that deep neural network approach outperforms multiple model adaptive approach in terms of accuracy and the amount of required measurements for the estimation, for all examined scenarios.