Improvements on matching pursuit based on fractional Fourier transform

The matching pursuit (NIP) algorithm is adaptive signal representation method. In this paper, we made improvements upon the NIP based on the fractional Fourier transform. Firstly, the three-parameter Chirp atoms with the scale, the rotation and the radial-shift, are proposed. And secondly, the searching strategy of the best atoms in NIP is presented. The simulation results have shown that the improved NIP not only reduced the size of the Chirp dictionary and economized the time of searching the best atoms, but also provided the most time-frequency information by thanking to estimating and grasping the time-frequency structure in advance.