Instantaneous heart rate and cardiac dynamics can be assessed through multiple noninvasive physiological signals that reflect different biophysical principles: electrical activity (ECG), acoustic vibrations (PCG), chest motion (SCG), and volumetric or mechanical changes at the periphery (PPG, PiPG). While individual properties of these modalities are described in the literature, direct cross-modal comparisons recorded with synchronized hardware are scarce. This study addresses that gap by simultaneously acquiring ECG, PCG, SCG, PPG, and PiPG under standardized resting conditions and exporting their temporal and spectral content into a single comparative framework. The goal is to provide practical benchmarks for filtering, feature extraction, and multimodal cardiac monitoring design rather than to validate specific hardware implementations.
The experiment used a custom acquisition system developed at CICIT 1403 (Lille, France) to record five modalities synchronously at a 1000 Hz sampling rate. The setup was designed to align every modality on the ECG R-peak, enabling consistent time-domain comparisons and spectral analyses without inter-system bias introduced by asynchronous recordings or different front-ends.
The recorded modalities and the sensors used were:
ECG: Surface Ag/AgCl adhesive electrodes positioned in a standard Lead I configuration. A driven-right-leg (DRL) circuit and battery operation were used to minimize common-mode and power-line interference.
PCG: An electret microphone (high sensitivity, omni-directional) with an audio preamplifier providing wide bandwidth and low input noise captured heart sounds related to valve closure.
SCG: A high-resolution three-axis accelerometer was mounted on the lower sternum; the analysis focused on the Z-axis corresponding to the chest-axis vibrations related to cardiac motion.
PiPG: A piezoelectric sensor placed on a finger adjacent to the PPG probe measured local mechanical changes induced by arterial pressure waves.
PPG: A reflective fingertip sensor (infrared channel analyzed) measured blood volume changes at the fingertip.
All analog signals except the PPG channel were digitized using a 24-bit analog front-end at 1000 Hz to preserve fine temporal and spectral detail.
Synchronous digitization at the same sampling frequency allowed direct alignment of modalities on ECG R-peaks. The recording chain and battery-powered acquisition were intended to reduce external electrical noise and to keep system-related differences minimal, supporting fair cross-modal comparisons.
Signals were temporally aligned to the ECG R-peak to examine morphology and inter-modality delays. The study reports consistent temporal latencies that reflect known physiological conduction and mechanical propagation delays: electrical depolarization (ECG) precedes mechanical events detected in PCG and SCG, and peripheral volumetric/mechanical signatures in PPG and PiPG follow those central events. The recorded latencies provide an empirical map for matching events across modalities when performing multimodal fusion or when extracting interval-based features.
Frequency analysis across modalities identified modality-specific dominant bands and differing spectral richness. Key findings include:
PCG: A dominant band centered near ~30 Hz, consistent with the frequency content of cardiac sounds produced by valve closures.
SCG: A primary band reported in the ~11–14 Hz range, reflecting chest acceleration components associated with mechanical cardiac motion.
PPG and PiPG: Dominant energy in the low-frequency band around 1–2 Hz, corresponding to heart rate and its low-frequency harmonics in peripheral volumetric and mechanical signals.
ECG: Reproducible morphology and spectral features well documented in prior literature; retained high spectral fidelity in these recordings.
Bandwidth analyses highlighted that modalities differ markedly in spectral richness: ECG, PPG, and PiPG show reproducible spectra, whereas PCG and SCG exhibit broader variability in spectral distribution due to sensor coupling and anatomical factors.
By exporting time- and frequency-domain features into a single framework, the study demonstrates reproducible morphology and spectra for ECG, PPG, and PiPG, while PCG and SCG show higher inter-recording variability. The latter variability is attributed to factors such as sensor placement, anatomical coupling, and the acoustic/mechanical transfer functions of the chest and sensor interfaces. These differences influence the choice of preprocessing (for example, filter cutoffs), feature extraction windows, and the reliability of event detection across modalities.
The modality-specific bands and latency maps serve as practical guidelines for designing filters and feature extraction pipelines. For example, filtering strategies can target the identified dominant bands (PCG ~30 Hz, SCG ~11–14 Hz, PPG/PiPG ~1–2 Hz) while preserving relevant harmonics. Temporal alignment to ECG R-peaks is recommended when integrating mechanical, acoustic, and volumetric signals to account for physiological delays. The unified dataset and reference interpretations can inform sensor fusion algorithms, help select appropriate sampling and anti-aliasing strategies, and support the optimization of wearable cardiac monitoring systems.
A minimal dataset supporting the analyses is provided as Supporting_Information_file.xlsx. The study emphasizes that its objective is a signal-processing and physiological benchmark using identical acquisition hardware rather than a validation of particular commercial devices. Observed variability in PCG and SCG should prompt caution when generalizing sensor-specific performance without accounting for placement and coupling conditions.
Simultaneous recordings of ECG, PCG, SCG, PPG, and PiPG under standardized resting conditions produce a consistent comparative map of temporal latencies, modality-specific frequency bands, and relative spectral richness. These benchmarks are intended to guide filtering, feature extraction, and multimodal fusion in wearable and ambulatory cardiac monitoring. The published minimal dataset can be used as a reference for future studies and for development of integrated sensing systems.