Abstract
Accurate detection of cardiac events can improve cardiovascular monitoring and diagnostics capabilities. Specifically, detecting the fiducial points related to aortic valve opening (AO) and closure (AC) facilitates the extraction of important timing characteristics that are informative of ventricular performance. Seismocardiography (SCG) offers a noninvasive way to observe these mechanical cardiac activities; however, noise from motion or speech artifacts often compromises the reliable detection of these fiducial points, especially during stress testing conditions. Stressors such as exercise are critical in determining cardiovascular health, as they can reveal abnormalities not apparent at rest. This is particularly true for conditions like heart failure, where cardiac function may deteriorate significantly under stress. SCG provides a valuable tool to noninvasively monitor these mechanical changes. Stress testing combined with SCG may help identify early signs of heart failure. In this study, we propose a Kalman filter-based methodology, named Heart Rate Informed Kalman Filter (HIKAF), that leverages heart rate (HR) to robustly identify AO and AC using SCG, even under moderately noisy conditions. We compare its effectiveness against existing approaches through structured experiments across various physiological states. HIKAF achieves significant correlations with manual annotations, yielding Pearson's r values of 0.934 and 0.899 for the relative shifts in AO and AC points between the baseline and exercise stages, substantially outperforming existing algorithms. These results highlight that HIKAF adapts effectively to dynamic changes in cardiac mechanical events and remains robust under noisy conditions, offering promising potential to improve real-time cardiovascular monitoring in wearable devices, remote settings, and clinical applications.