GRANULAR REPRESENTATION OF BIOMEDICAL SIGNALS USING NUMERICAL DIFFERENTIATION METHODS
M. Momot, Alina Momot, Adam Gacek · Journal of Medical Informatics & Technologies · 2010
This work presents the general idea of granular des cription of temporal signal, particularly biomedica l signal sampled with constant frequency. The main idea of p resented method is based on using triangular fuzzy numbers as information granules in temporal and amplitude spac es. The amplitude space contains values of first fe w derivatives of underlying signal. The construction of data granule s is performed using the optimization method accord ing to some objective function, which balances the high coverag e ability and the low support of fuzzy numbers. The granules (descriptors) undergo the clustering process, namel y fuzzy c-means. The centroids of created clusters form a granular vocabulary and the quality of description is quanti tatively assessed by reconstruction criterion. There are presented results of experiments with the electrocardiographic signal, digitally sampled and stored in MIT-BIH database. The method of numerical differentiation of function based on finite set of its value s is employed, which incorporates polynomial interpolation. The pa per presents results of numerical experiments which show the impact of method parameters, such as temporal window length, degree of polynomial, fuzzification param eter, on the reconstruction ability of presented method.