Millimeter wave target identification system using a backpropagation neural network

An integrated radar target identification system for millimeter wave radar measurements has been developed using a neural network approach. The main objective was to achieve target identification independently of the target position, size and aspect angle. Measurements of simple and complex targets have been made using a 94 GHz millimeter wave FM-CW radar. Simulations have been used to validate the algorithms. Feature vectors extracted from range signatures and Doppler signatures are fed into a multi-layer neural network which uses the backpropagation training algorithm for the learning phase. The accuracy of the identification achieved is discussed in view of the results expected. The complexity required to generalize this approach to more complex targets is examined