Customized ECG beat classifier using mixture of experts

Reports an effort to develop a user-adaptation (customization) algorithm which, with minimal human supervision, is capable of adapting the performance of an ECG beat classification algorithm to the special characteristics of individual patients. In the authors´ algorithm, they assume a black box ECG beat classifier, and a small set of user-specific training data, edited by human experts, are given. The objective is to devise a method to take advantage of the small user-specific training data set to achieve high classification rate rather than using the black box ECG beat classifier alone