OPTIMIZED OVERCOMPLETE SIGNAL REPRESENTATION AND ITS APPLICATIONS TO TIME-FREQUENCY ANALYSIS OF ELECTROGASTROGRAM

The electrogastrogram (EGG) is a surface measurement of gastric myoele ctrical activity. The normal frequency of gastric myoelectrical activi ty in humans is 3 cycles/min. Abnormal frequencies in gastric myoelect rical activity have been found to be associated with functional disord ers of the stomach. The aim of this article was, therefore, to develop new time-frequency analysis methods for the detection of gastric dysr hythmia from the EGG. A concept of overcomplete signal representation was used. Two algorithms were proposed for the optimization of the ove rcomplete signal representation. One was a fast algorithm of matching pursuit and the other was based on an evolutionary program. Computer s imulations were performed to compare the performance of the proposed m ethods in comparison with existing time-frequency analysis methods. It was found that the proposed algorithms provide higher frequency resol ution than the short time Fourier transform and Wigner-Ville distribut ion methods. The practical application of the developed methods to the EGG is also presented. It was concluded that these methods are well s uited for the time-frequency analysis of the EGG and may also be appli cable to the time-frequency analysis of other biomedical signals. (C) 1998 Biomedical Engineering Society.