Amplitude Rescaling Influence on QRS Detection

Abstract

When we record the electrical activity of the heart we generate a signal called an electrocardiogram. Within the electrocardiogram, the information that explains the heart’s health is based on the detection of QRS complexes. The focus of this paper is on a wearable ECG sensor that uses a low sampling frequency and bit resolution while it converts the analog signal to digital data. The overall goal is to see if an efficient industrial QRS detector can be developed within these constraints. In particular, we set a research question to investigate how amplitude rescaling affects sensitivity and positive predictive rate of the Hamilton algorithm for QRS detection and improved it by optimizing it based on amplitude ranges. We used the MIT-BIH Arrhythmia ECG database to evaluate performance. The original recordings are sampled with a sampling frequency of 360 Hz with a 11-bit resolution over a 10 mV range. Our experiments include testing rescaled signals on a sampling frequency of 360 Hz using different maximum amplitudes. We found that rescaling impacts performance and that the optimization parameters need to tuned to obtain the expected performance. However, the performance decreases when the maximum amplitude is lower than 9 bits.