A characteristic points unified approach for ECG analysis and compression

Characteristic points detection in ECG signal for analysis and compression has been tackled mostly by sequential strategies that take into account only local properties of such features. We propose a recursive strategy that takes into account a segment wise global properties of such features. The approach uses no filtering tools and is rather exclusively based on pattern recognition tools. We use segment wise recursively computed characteristic points as the main tool for compression and smoothing of ECG data through linear interpolation between successive such points. A classification process of the computed characteristic points yields baseline correction and main waves extraction. A points of confidence discrimination process is used for R waves detection. Intensive tests show the method effective usefulness as a unified approach for ECG processing.