Development of A Novel Noise Detection Algorithm Using Attractor

Abstract

We propose a novel method of recognizing a noise. In recent years, the electrocardiogram (ECG) is widely used not only in the hospital but also in home care and health management situation. However, the measurement of ECG is susceptible to motion noise or other external factors. The noise can easily degrade the performance of automatic analyzers and can cause false alarms. In this study, we quantify an irregular disorganized pattern of the noises using attractor analysis. The ECG signal is converted to the two-dimensional time-delay coordinates, and the internal area of the attractor is split up into equal intervals. The irregularity index is defined as the products of the correlation coefficient and the entropy of each area. This index is evaluated using the MIT-BIH arrhythmia database. The result suggests the proposed method is able to classify the ECG or noises.