A Microwave Photonic Radar Warning Receiver based on Deep Compressed Sensing

A wideband photonics-based radar warning receiver that exploits a novel deep-neural-network-enabled compressed sensing technique is proposed and experimentally demonstrated. In the scheme, first, the radar input signal is mixed with a pseudo-random-sequence electrical signal using a wideband microwave photonic mixing approach; then, the mixed signal is filtered with a low-pass electrical filter. The resulting compressed signal is digitized using a low-speed analog-to-digital converter. Instead of relying on a computationally expensive compressed sensing waveform-recovery algorithm, followed by conventional parameter-extraction processing, a deep neural network algorithm is used to rapidly estimate the radar waveform parameters directly from the low-frequency compressed signal. A proof-of-concept receiver is tested in a laboratory experiment using pulsed radar signals with different center frequency, bandwidth, and pulse repetition frequency: it reaches an analog input bandwidth of 5 GHz, limited by the available laboratory equipment, a compression factor of 20 (i.e., the digitized signal has frequency content in the 0-250 MHz range), very accurate estimation of the radar waveform parameters, and processing time of a few milliseconds.