Efficient FPGA Implementation of a Convolutional Neural Network for Radar Signal Processing

Although neural networks, especially convolutional neural networks (CNNs), have been successfully applied to many domains, there have not found many radar applications mainly due to a paucity of available training data. Focusing on fixed-site radars, this work uses in-situ collected data to train a CNN classifier and suppress clutter components that allow targets to “hide in plain sight.” This paper describes a software and hardware co-design approach for implementing a neural network to improve radar signal processing. At the algorithm level, we propose using the ResNet10 model structure and other optimizations trained using the angle-Doppler spectrum of returns at each range. The FPGA implementation is then carefully optimized to better tradeoff performance and energy efficiency. Experimental results show our approach achieves better performance than conventional methods and exceed the requirement by more than 2.5×. Meanwhile our energy consumption is much lower than other platforms like GPU. Our optimization methods can be applied to other CNN structures for efficiency improvement.