Signal enhancement with matching pursuit

Matching pursuit is a greedy algorithm that adaptively decomposes any signal into a linear weighed expansion of atomic vectors that are selected from a highly redundant dictionary of vectors. We study signal enhancement with matching pursuit. By tracing and analyzing the procedure of matching pursuit with a specific Gabor dictionary, we find a way to obtain the underlying signal from noise with a coherent ratio threshold, as well as a method of determining the coherent ratio threshold. It is proved by experiments that the method can efficiently enhance a signal corrupted by noise without prior information about the noise and the signal. Finally, many test results are given, which are generally better than those of wavelet denoising.