Comparison of Fast Fourier Transforms with Other Transforms in Signal Processing for Tactical Radar Target Identification.

摘要

The High Resolution Radar Branch of the Rome Air Development Center has developed a tactical target identification (TTI) pulsed-Doppler radar system which generates two-dimensional 'images' of aircraft. The signal processing technique utilizes the fast Fourier transform (FFT) to produce a slant-range versus cross-range display. If the TTI system is to be effectively employed in an aerial warfare environment then real-time processing is necessary. In an effort to speedup the signal processing several alternative transforms were studied as possible substitutes for the FFT. The Karhunen-Loeve, Cosine (Sine), Mellin, and Hankel transforms were investigated and found to be infeasible for use in TTI imaging. The Walsh (Hadamard) transform was studied in detail and tested in a simulation program and found that it could not be utilized in the TTI signal processing. Two methods of converting from the Walsh sequency domain to the Fourier frequency domain were studied. The first scheme, a recursive relationship between the arithmetic and logical autocorrelation functions as presented by Robinson was discovered to be incorrect. The second, a method of computing the Fourier coefficients from the Walsh coefficients of a function was demonstrated to be too time consuming to be implemented in TTI signal processing. Several floating-point FFT implementations were tested using the simulation program. Also, several fixed-point FFT algorithms were derived and tested. All of these were evaluated on the basis of speed and memory requirements and one fixed-point FFT algorithm was shown to be fast enough and accurate enough for implementation on the TTI Minicomputer. (Author)