Development of a real-time QRS beat classifier using a nonlinear trimmed moving averaging-based system

In this paper, a real-time QRS beat classification system devised based on a nonlinear trimmed moving averaging filter is presented. Such a nonlinear system aims to identify the abnormal beat of ventricular origin from the normal one. The proposed beat classifier can function in parallel with a real-time QRS detector, permitting the tasks of beat detection and diagnosis to alternate with each other. Algorithm performance was evaluated against the ECG recordings selected from the MIT-BIH arrhythmia database. Numerical results demonstrated that over 99.8% beat identification rate can be achieved by the algorithm.