Sensor Technologies for Wearable and Portable Cardiac Acoustic Monitoring: An Experimental Study

Phonocardiography is a promising technique for continuous cardiovascular monitoring, and recent advances have extended its use beyond conventional clinical environments toward portable and wearable applications; however, selecting a suitable sensor for these systems remains challenging because different sensing technologies involve distinct trade-offs in performance and implementation. This study compared five sensing technologies: Electret and microelectromechanical systems (MEMS) microphones, a contact microphone, a polyvinylidene fluoride (PVDF) film sensor, and a capacitive MEMS accelerometer. Bench characterization assessed frequency response, current consumption, and output-referred voltage-noise density. PCG recordings were obtained from 10 healthy volunteers at rest and after physical activity. Mixed-effects models analyzed recording-level median signal-to-noise ratio (SNR) from 85 of 100 recordings meeting the original joint S1/S2 evaluability criterion, with sound amplitude and noise measured on signals band-pass filtered at 20-150 Hz. Sensor type had a statistically significant overall effect on S1 (p < 0.001) and S2 SNR (p = 0.0025). Holm-adjusted comparisons identified higher S1 SNR for PVDF and the contact microphone than for the Electret microphone and accelerometer, and higher S2 SNR for PVDF and the contact microphone than for the accelerometer. S1 SNR increased post-activity by an estimated 2.74 dB (95% CI: 0.93-4.55 dB); the condition effect for S2 was not statistically significant. Across two subsequent visual assessments of all recordings, S1 detectability was 84-89%, and S2 detectability was 54-56%, with exact intraobserver agreement of 89% and 64%, respectively. These findings identify sensor differences within the originally evaluable subset, while variable detectability and annotation repeatability limit broader performance claims.