Matching pursuits with time-frequency dictionaries-英文文献.pdfVIP

IEEE TRANSACTIONS ON SIGNAL PROCESSING. VOL. 41. NO. 12, DECEMBER 1993 3397 Matching Pursuits With Time-Frequency Dictionaries Stephane G. Mallat, Member, IEEE, and Zhifeng Zhang Abstract-We introduce an algorithm, called matching pur- ideas, it requires to use circumvolutions that replace un- suit, that decomposes any signal into a linear expansion of available words by full sentences. waveforms that are selected from a redundant dictionary of Flexible decompositions are particularly important for functions. These waveforms are chosen in order to best match the signal structures. Matching pursuits are general proce- representing signal components whose localizations in dures to compute adaptive signal representations. With a time and frequency vary widely. The signal must be ex- dictionary of Gabor functions a matching pursuit defines an panded into waveforms whose time-frequency properties adaptive time-frequency transform. We derive a signal energy are adapted to its local structures. Such waveforms are distribution in the time-frequency plane, which does not in- called time-frequency atoms. For example, impulses need clude interference terms, unlike Wigner and Cohen class dis- tributions. A matching pursuit isolates the signal structures that to be decomposed over functions well concentrated in are coherent with respect to a given dictionary. An application time, while spectral lines are better represented by wave- to pattern extraction from noisy signals is described. We com- forms which have a narrow frequency support. When the pare