Multisource Signal Fusion using Dempster Shafer Evidence Accumulation Concept and its Applications to CMFD and Multimodal Biomedical Image Fusion

This paper addresses a soft computing approach of fusion of signals from different independent sources. The signals may be from different types of primary classifiers. The Dempster Shafer Evidence Accumulation (DSEA) theory provides a robust platform for evidence fusion and it incorporates uncertainty, imprecision and conflicting situations in the process of decision making into a mathematical framework. Primarily, Neuro-Fuzzy classifiers have been used on the signals of each individual source to classify them into meaningful clusters and to assign mass value to each cluster, then Dempster Shafer Evidence Accumulation engine (DSEAE) has been used to combine them for final output with proper classification to different admissible clusters. We have cited two experimental results of the use of this concept. Firstly, the concept has been studied on a diesel engine to fuse the coolant flow signals from three primary ANN classifiers; secondly, it has been used to fuse SPECT and MR-T2 registered brain images classified by fuzzy C-means method.