Survey on Electrocardiogram Signals Analysis and Characterization

Abstract

The analysis and evaluation of Electrocardiogram (ECG) signal quality plays a crucial role in significantly enhancing the diagnostic accuracy and reliability of unsupervised ECG analysis systems. In practical scenarios, ECG signals are frequently subject to various types of noise and artifacts. Consequently, numerous assessment methods have been introduced, utilizing features from both the ECG signal and noise alongside machine learning classifiers or heuristic decision rules. This article presents a survey on current state-of-the-art methods and underscores the practical limitations of existing assessment approaches. Drawing from previous studies and our own research, it is evident that there is a substantial demand for a lightweight ECG noise analysis framework. Such a framework is essential for real-time detection, localization, and classification of single and combined ECG noises, especially within the constraints of wearable ECG monitoring devices that often have limited resources.