Efficient Ordinal Pattern-Based Encoding for Data Analysis: Application to ECG Beat Classification

Complexity measures are useful for the analysis of a system as a whole and allow to detect system malfunctioning. They apply to biology, geophysics, medicine, mechanics, and geography, to cite a few. As a complexity measure, ordinal pattern-based algorithms are efficient for discriminating time series, especially due to their high robustness against noise. However, the observation time they require prevents them from efficient application to experiments with limited data lengths. An example is given for low-resolution ECG beat analysis where the beat length does not exceed 500 samples. The proposed project aims to develop a new complexity measure based on an optimized ordinal pattern-based encoding approach to enhance the observation time requirement. The efficiency of the proposed approach will be verified in the detection and classification of ECG beats with premature ventricular and atrial contractions. Annotated data from Physionet will be used for validation purposes.