KRIGING-BASED POSSIBILISTIC ENTROPY OF BIOSIGNALS

This paper presents an approach for nonlinear dynamical analysis of complex time-series data using the principles of the approximate entropy family, geostatistics, and possibility. Uncertainty of the measure of signal similarity is modeled using the concept of fuzzy sets and quantified by the signal error matching. The proposed method has the ability to discern the signal complexity at a more detailed level than the approximate entropy as well as to incorporate the spatial information inherently existing in the signal characteristics. Based on experimental results on the study of mass spectrometry data for cancer study, the proposed method appears to be a promising tool for classification of biosignals.