An Overview of ECG Signal Processing and Analysis Techniques for Categorization of Cardiac Diseases

An essential diagnostic technique for assessing cardiac health is an Electrocardiogram (ECG). The heart's electrical activity is captured in this recording. The need to share the workload among physicians and relieve pressure on them has led to the development of automatic detection and classification techniques for heart arrhythmias and other abnormalities as the number of heart patients has increased. All detection and classification techniques operate in the following stages: signal preprocessing, which includes denoising, extracting features, and categorising features. Recently, several methods have been used to denoise, extract features, and categorize ECG signals. The preprocessing of the ECG signal is necessary before the extraction phase because numerous noise sources in a medical setting can deteriorate the signal. The present study reviews ECG signal analysis, feature extraction, and denoising techniques. Frequency domain filters adaptive, and Wavelet Transform (WT) based filters are commonly used to denoise ECG signals. For the ultimate classification task, various morphological, temporal, and statistical features, Fourier transform, and wavelet-based coefficients are frequently extracted from the ECG signals. Findings show that deep learning methods are best among the others for the classification task and that hybrid features increase detection efficacy. Most authors have attempted to categorize ECG into five classes. There is scope to identify the features that combine most effectively to provide better performance in categorising more heart diseases. Also, there is a scope for developing a classifier that performs better to classify a more significant number of heart arrhythmias or diseases.