Model-based sub-Nyquist sampling and reconstruction technique for ultra-wideband (UWB) radar

Abstract

The Army Research Lab has recently developed an ultra-wideband (UWB) synthetic aperture radar (SAR). The radar has been employed to support proof-of-concept demonstration for several concealed target detection programs. The radar transmits and receives short impulses to achieve a wide-bandwidth from 300 MHz to 3000 MHz. Since the radar directly digitizes the wide-bandwidth receive signals, the challenges is to how to employ relatively slow and inexpensive analog-to-digital (A/D) converters to sample the signals with a rate that is greater than the minimum Nyquist rate. ARL has developed a sampling technique that allows us to employ inexpensive A/D converters (ADC) to digitize the widebandwidth signals. However, this technique still has a major drawback due to the longer time required to complete a data acquisition cycle. This in turn translates to lower average power and lower effective pulse repetition frequency (PRF). Compressed Sensing (CS) theory offers a new approach in data acquisition. From the CS framework, we can reconstruct certain signals or images from much fewer samples than the traditional sampling methods, provided that the signals are sparse in certain domains. However, while the CS framework offers the data compression feature, it still does not address the above mentioned drawback, that is the data acquisition must be operated in equivalent time since many global measurements (obtained from global random projections) are required as depicted by the sensing matrix Φ in the CS framework. In this paper, we propose a new technique that allows the sub-Nyquist sampling and the reconstruction of the wide-bandwidth data. In this technique, each wide-bandwidth radar data record is modeled as a superposition of many backscatter signals from reflective point targets. The technique is based on direct sparse recovery using a special dictionary containing many time-delayed versions of the transmitted probing signal. We demonstrate via simulated as well as collected data that our design offers real-time (with single observation as oppose to equivalent-time with many observations) data acquisition of the wide-bandwidth radar signals using the sub-Nyquist sampling rate.