Application of fuzzy neural network to ECG diagnosis

In automated diagnosis of electrocardiographic (EGG) signals, the conventional approach requires a human expert to formulate the rules and has not the ability of learning. A neural network is capable of learning, but it needs large training sets, and its training speed is too slow. In addition, we can not understand thoroughly the knowledge that follows the changes of hidden neurons. All these limits hamper the practical application of neural network to ECG diagnosis. In this study, a fuzzy neural network (FNN) is designed to translate directly the expert knowledge into the neural network structure by using a fuzzy model. Thus, it can be applied to a diagnostic system without training process, and learn from the responding data in the running to improve the accuracy of the system. Thus, it possible to use FNN to interpret the knowledge that hides in the neural network