High PSL Radar Pulse Compression using Neural Network

Modern radar systems that operate in noise, strong clutter and jamming require advanced signals to help in processing of target echoes. One of the inevitable feature of such advanced radar signals is pulse compression which is responsible for the accuracy, resolution and determination of ambiguity of the range of the target. In the process of maximizing SNR by a factor of the time-bandwidth product of the waveform by a matched filter, relatively strong sidelobes of fixed Peak-to-sidelobe(PSL) ratio remain in the autocorrelation function. Use of certain "mis-matched" filters help in reducing the sidelobes but at the cost of reduced resolution. In this paper, an approach using a multilayered feed forward neural network for pulse compression is detailed for its near zero level of sidelobes also without compromising on the range resolution of the radar.