| When dealing with "Doppler effect" radar, the frequency information of the processed signals is extremely important. It is thus necessary to have an effective method. Moreover, the quantity of data to process is huge, so our algorithm must be fast. Since several years, the solution to this problem was the windowed Fourier transform1. The transform is indeed fast and (in most cases) effective if properly used.;As we work in a specific domain, all the processed signals do have some common characteristics. This characteristics make the matching pursuit (in is actual state) a nonmarketalbe solution to our problem. We then study the possibility of using the chirped matching pursuit [21, [171, [181, this algorithm seems to be better adapted to the characteristics of our signals.;Finally, as the software must be fast to be marketable, we introduce a modification to accelerate the processing time needed by the matching pursuit. These modifications will also make the matching pursuit better adapted to our signals characteristics.;1It is also called short time Fourier transform;However, the windowed Fourier transform suffer of a flexibility problem. On other hand, the size of the window cannot be changed in one analysis. It is the main problem of this work. A solution to this weakness is the matching pursuit algorithm [16l. We think that in most of the cases, it can help to eliminate the weaknesses of the windowed Fourier transform. |
