Applicability of qualitative ECG processing to wearable computing

Studies of ECG time-series properties and complexities are significant part of the research on the possibilities to automate ECG classification by a wearable body computer. Numerous statistical measures, as well as more recently introduced non-linear and complexity measures provide the basis for signal classification, prediction of events, and discovery of underlying systems and models expressing the observed heart dynamics. This paper presents qualitative signal discretization, based on persistent state trend definition. This transformation results in a compact symbolic sequence representation of the original time series. The information content of the transformed sequence is assessed using some of the classic signal complexity and similarity measures, adapted to the new representation. The presented methodology is applied to ECG time signals classification.