A modified fast matching pursuit algorithm based on signal energy property

Decomposition of signals over a series of time-frequency atoms by the Matching Pursuit (MP) algorithm has found many applications in signal processing and harmonic analysis. But traditional MP algorithm, such as the method based on inner product lib, are not able to handle high-dimension signal in a reasonable memory requirement and computational time. Benefit from the best time-frequency resolution of Gabor atoms, through building new dictionaries from the over-complete dictionary exploiting the structure of the signal energy property, a novel fast MP algorithm which can reduce computation and storage requirement dramatically is proposed to solve this problem. Simulation shows that the speed of the novel algorithm is more than 100 times faster than that of the traditional MP algorithm and verifies the effectiveness of the proposed algorithm.