Application of state space frequency estimation techniques to radar systems

Many applications for FMCW radar systems require the resolution of several closely spaced frequencies. In general the Fast Fourier Transform (FFT) is used for spectrum estimation, despite the inherent resolution problems. To overcome the resolution limitation the state space approach has been proposed by Rao and Arun (1992), Marple (1988) and Kay (1988). However, most experimental results in the above papers are based on simulations. This paper points out the differences of the "real world" signal structure to simulated signals, and describes the problems posed by real radar signals. For one of the key steps of the state space algorithm, model order selection, a novel algorithm based on a posteriori analysis is introduced. The feasibility of the new approach is verified with an actual 24 GHz level gauging FMCW radar system. Our approach yields stable and accurate results with a resolution approximately three times higher than the FFT resolution