A polycoherence-based ECG signal non-linearity detector

Abstract

This paper proposes a new algorithm to detect non-linearities in ECG signals and to determine the order of non-linearity. The ECG bispectrum is analysed and the bicoherence index is calculated to identify non-linearity. The diagonal slice of the polycoherence index of any order, calculated using the diagonal slice of the polyspectrum of the same order and the power spectrum, is proposed as an estimator. The possibility of higher-order non-linearities in ECG signals is investigated using these slices. Physiological and pathological cases have been studied. The analysis of the bispectrum and bicoherence index reveals tight quadratic interactions among frequencies in the region (0–20 Hz), which suggests the existence of non-linearities. The polyspectrum and polycoherence slices indicate the presence of higher-order phase-coupled harmonics, in the physiological cases, which is attributed to higher order non-linearities. Differences between physiological and pathological cases are assessed and a decrease in the non-linearity order could be correlated with pathological conditions.