Modeling radar data with time series models

In Frequency Modulated Continuous Wave (FMCW) radar, the Fast Fourier Transform (FFT) is very efficient in separating reflections from various range cells. For the subsequent determination of the Doppler spectrum, the FFT is less suitable. The raw FFT is very erratic, while the smoothed FFT is limited in the spectral shapes it can accurately describe. For determination of the Doppler spectrum, ARMAsel time series analysis is better than the FFT. With ARMAsel time series analysis, an ARMA model is estimated from the data. The Doppler spectrum can be calculated from the ARMA-parameters. The model order and type are determined automatically from the data. Improved accuracy in comparison to other time series techniques is obtained by using robust estimation algorithms and order selection criteria, and by selecting automatically between several model types.