ECG: an electrical timing reference AI
An electrocardiogram (ECG) records electrical activity associated with the heartbeat. Its prominent QRS complex provides a convenient timing reference for aligning other signals. Electrical activation is different from the mechanical movement measured by BCG and the peripheral pulse measured by PPG. The American Heart Association explains the signal’s clinical role.
How it is used here
An offline detector filters ECG into a QRS-sensitive band, computes derivative energy, integrates it over 150 ms, identifies candidates and refines their location against the original ECG. This is inspired by Pan–Tompkins; it is not an exact implementation or manual annotation. Detected fiducials define the heartbeat windows used throughout this project.
Those windows align reference pressure, PPG and BCG measurements. ECG also contributes waveform descriptors to the extended model. ECG morphology is measured from −0.2 to +0.8 cycle durations around the fiducial so that the leading complex is not truncated.
A mistimed or missed fiducial can change several downstream features at once. The plotted markers are approximate, and waveform descriptors are generic measurements rather than clinical ECG intervals. For the measurements, see waveform features; for complementary signals, see PPG and BCG.