Matching pursuit algorithm and application in MI-EEG
Ying Long Zhou · Journal of Circuits and Systems · 2005
This paper present the basis principle of matching pursuit algorithm based on stochastic time-frequency dictionary, and its application to motor imaginary electroencephalogram (EEG). The method can decompose signal into linearity combination of a series of time-frequency atoms, and randomizes the parameters of atoms in the dictionary before each decomposition, which has high time-frequency resolution and parametrization of signal's microstructure. Simulated analyzing demonstrates that most of the basic time-frequency characteristics can be represented even in low SIN ratio. For actual motor imagination EEG data, it is found that some rules of special frequency bands, such as alpha beta, are consistent with the mu rhythm about motor sense, moreover they appear event-related EEG changes.