Detecting Characteristic Points for the Analysis of Bioimpedance Signal Through a Synergy of Fuzzy Rule-based models and Granular Neural Networks

In this study, we propose a novel methodology for determining accurate positions of characteristic points encountered in the analysis of bioimpedance signals. The proposed approach fully utilizes two fundamental modeling pursuits based on fuzzy rule-based models and neural networks. We take advantages of the unique capabilities of fuzzy rule-based models to characterize the nonlinear relationship between the acquired bioimpedance signals and its temporal coordinate. The fuzzy modeling approach is used to approximate the process that generates the bioimpedance signals through a collection of rules (if-then statements). In the sequel, the parameters of the models are used as the inputs of neural network models to determine the position of characteristic points. We further augment the numeric neural network to its granular counterpart to accommodate the uncertainty in the available experimental evidence through allocating a certain level of information granularity across the parameter space. The resulting granular outputs (intervals) become reflective of the quality and a level of confidence associated with the prediction results. The quality of the prediction results is quantified in terms of the coverage and specificity criteria. The performance index is also enhanced to deal with the situation when the positions provided by experts are also information granules (intervals). The performance of the proposed approach is justified through a collection of experiments carried out on the collected real-world bioimpedance signals. Experimental results show that the proposed approach achieved higher accuracy in determining the position of characteristic points in comparison with other existing methods.