Extracting input features and fuzzy rules for detecting ECG arrhythmia based on NEWFM

Abstract

Fuzzy neural networks have been successfully applied to generate predictive rules for medical or diagnostic data. This paper presents an approach to automatically detect ECG arrhythmias using the neural network with weighted fuzzy membership functions (NEWFM). NEWFM classifies normal and abnormal beats by the trained bounded sum of weighted fuzzy membership functions (BSWFMs) using normalized features in the range of [0, 1] from UCI repository of machine learning. The generalized 4 features, locally related to the time signal, are extracted by the non-overlap area measurement method. The total numbers of samples are 452 data. The 80% of the data are used for training and 20% for testing. The result of accuracy rate is 81.32%. The BSWFMs of the 4 features trained by NEWFM are shown visually, which makes the features interpret explicitly. Since each BSWFM combines multiple weighted fuzzy membership functions into one using bounded sum, the four small-sized BSWFMs can realize real-time ECG arrhythmias detection in mobile environment.