Identification of R-peak occurrences in compressed ECG signals

Heart Rate (HR) is one of the mostly used electrocardiogram (ECG) feature in many automatic detectors of anomalies. This paper deals with a preliminary study on a novel approach which, through the combination of Machine Learning (ML) and Compressed Sensing (CS), aims at retrieving vital information from a digital compressed single-lead electrocardiogram (ECG) signal. As a potential key information to estimate the heart rate, this study focuses on the identification of R-peak occurrences. The study has been conducted on two different types of signal both obtained from the compressed samples provided by a CS algorithm, already available in literature. The results demonstrate that the use of CS in combination with a ML technique can find high competitiveness when compared to a state of the art method working on the uncompressed ECG signal.