Key result
Acoustic resonator with neural network non-invasively reconstructs ECG with ~4 BPM heart rate error.
Why the study?
Non-invasive methods for accurate heartbeat sensing and ECG waveform reconstruction are needed for prognostic health monitoring and cardiac biomonitoring in dynamic environments.
Does an acoustic Helmholtz resonator cavity with neural network reconstruction accurately measure heart rate and R-peak timing compared to standard ECG in seated human volunteers?
Does an acoustic Helmholtz resonator cavity with neural network reconstruction accurately measure heart rate and R-peak timing compared to standard ECG in seated human volunteers?
Effect estimate: HR error 4.23 ± 4.50 BPM; R-peak MAE 36.50 ± 53.17 ms
A novel acoustic Helmholtz resonator integrated into a seat and coupled with neural network processing can non-invasively reconstruct ECG waveforms with reasonable accuracy in a controlled, stationary setting.
May support contactless cardiac monitoring; hypothesis-generating for acoustic ECG reconstruction pending clinical validation.
Physiological markers are a primary source of prognostic health monitoring systems. Specifically, with heartbeat sensing, many ailments may be identified in their nascent stage. In clinical settings, the human heartbeat is measured by way of an electrocardiogram (ECG). Common metrics for heartbeat analysis include the heart rate (HR), or average beat-to-beat value over time, and heart rate variability (HRV), or variation of beat-to-beat intervals over time. In day-to-day cardiac monitoring, contact-based sensing techniques are realized using various wearable technologies. Here, we present the design, experimental testing, and proof-of-concept validation of a passive acoustic Helmholtz resonator cavity for heartbeat sensing in the [50-120] Hz frequency range and determination of inter-beat interval (IBI) in a non-invasive manner using sound. Integral to this sensing process, a unique neural network-based machine learning algorithm is exploited to reconstruct the ECG waveform from the resonator-enhanced acoustic signal. The system achieved a HR estimation error of 4.23 ± 4.50 BPM and an R-peak timing mean absolute error (MAE) of 36.50 ± 53.17 ms relative to a simultaneously acquired reference ECG. The potential extension of the engineered sensor hardware plus software system for future non-invasive cardiac biomonitoring of seated users in a dynamic vehicle environment is discussed. Paul Schmalenberg and colleagues present the design, testing, and validation of an acoustic Helmholtz resonator cavity for heartbeat sensing. A neural network-based algorithm accurately reconstructs the electrocardiogram waveform from the measured acoustic signal, which enables cardiac biomonitoring of seated users in dynamic vehicle environments.
No takes yet. Share an insight, caveat, or question.
Schmalenberg et al. (2026) studied Cardiac monitoring (n=13). Acoustic Helmholtz resonator with neural network ECG reconstruction vs. Standard contact-based ECG was evaluated on Heart rate estimation error and R-peak timing mean absolute error (HR error 4.23 ± 4.50 BPM; R-peak MAE 36.50 ± 53.17 ms). A passive acoustic Helmholtz resonator combined with a neural network reconstructed ECG waveforms non-invasively, achieving a heart rate estimation error of 4.23 ± 4.50 BPM compared to standard ECG.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: