Radar Behavior Classification Based on DBN

This article introduced the application of restricted Boltzmann machine in deep belief network. A revised algorithm of evaluation of reconstruction error is used to extract the different features of multi-mode radar in different work modes. We simulated radialation data of a multi-mode radar which has five work modes of different aims. First, we introduce and choose the main parameters of multi-mode radar signals to input to the DBN network. Wth the data used in deep belief network to train the model parameters, we got a well-performed model. By comparing with BP methods, our model has a higher accuracy rate than the former means.