R-peak detection based chaos analysis of ECG signal

Abstract

Electrocardiography (ECG) is a non-invasive test that is used for recording contraction and relaxation activities of the heart by using an electrocardiogram. Early detection of abnormalities of the heart through ECG is essential for reducing the prevalence of casualties due to cardiac arrests worldwide. In this study, physioNet ECG records have been considered for analysis. During recording, ECG signal is also affected by various noises, where analog filters fail due to the effect of temperature and drift, and digital filters fail due to inappropriate selection of passband and gain parameters. For adequate and frequent usage in the medical field, it demands correct and precise R-peak (QRS-complex) detection; which requires an appropriate combination of pre-processing, feature extraction and detection techniques. Therefore, independent component analysis (ICA) is used in the pre-processing stage due to nonlinear nature of the ECG signals and chaos analysis is applied for feature extraction for different ECG databases. The ICA method separates an individual signal from mixed signals by assuming that the original underlying source signals are mutually independently distributed. Chaos analysis examines the irregular attitude of the system and fits it into deterministic equations of motion. Chaos analysis is implemented by plotting different attractors against various time delay dimensions. R-peak detection is well known to be useful in diagnosing cardiac diseases. The R-peaks are detected using principal component analysis (PCA) which outperforms the existing state-of-the-art techniques.