Introduction

Managing hypertension and cardiovascular risk demands frequent and accurate blood pressure (BP) monitoring. Traditional cuff-based methods, however, are cumbersome and interrupt daily life. Our ICASSP 2025 paper—“Evaluation of Wearable Head BCG for PTT Measurement in Blood Pressure Monitoring”—presents a novel approach by leveraging a fully standalone True Wireless Stereo (TWS) or Over-the-ear Wireless System (OWS) earbud. In this system, the earbud is equipped with sensors to capture both ballistocardiography (BCG) and photoplethysmography (PPG) signals, enabling the measurement of artery pulse transit time (PTT) for cuffless BP estimation.

This blog post provides a comprehensive overview of our prototype design, experimental methodology, signal processing pipeline, and detailed results—all of which underscore the potential of integrating BP monitoring into everyday wireless earbuds.

Background and Motivation

Pulse Transit Time (PTT) for BP Monitoring

Pulse transit time (PTT) is defined as the time taken by the arterial pulse wave to travel from the heart (proximal timing) to a peripheral site (distal timing). Traditionally, PTT is obtained using a combination of an electrocardiogram (ECG) to mark the cardiac electrical event and photoplethysmography (PPG) at the finger to mark the pulse arrival. Because PTT inversely correlates with BP (i.e., higher BP leads to stiffer arteries and a shorter transit time), it has become a popular surrogate for continuous, cuffless BP tracking.

Ballistocardiography (BCG) and Its Transition to Ear-Worn Devices

Ballistocardiography (BCG) measures the subtle recoil of the body resulting from the ejection of blood with each heartbeat. Historically, BCG was recorded with large platforms (beds, chairs, or scales). Recent miniaturization of inertial sensors now enables BCG capture through wearable devices. Our approach takes advantage of these developments by embedding an accelerometer and an optical sensor within a TWS/OWS earbud. By doing so, we can capture:

The dual-sensing capability allows a single, standalone device to compute PTT and, after calibration, estimate BP trends continuously.

System Design and Prototype

Unlike previous head-mounted approaches, our design leverages the familiar and comfortable form factor of TWS/OWS earbuds. The device integrates:

This design minimizes the number of sensors needed on the body and makes continuous monitoring feasible without sacrificing everyday comfort.

The Figure in below is the experiment procedure overview, showing the sequence of REST, Intervention, and Plateau sessions during the leg press task.

Figure 1. Experiment procedure overview, showing the sequence of REST, INT, and PLAT sessions during the leg press task.

Experimental Protocol

Participant and Data Acquisition

Our study involved human participants under controlled conditions. The experimental protocol consisted of three phases:

Simultaneously, the earbud captured BCG and PPG signals while a reference impedance cardiography (ICG) system recorded chest signals. The ICG served as the ground truth for cardiac events (e.g., the aortic valve opening).

Data Collection

Data were acquired in synchronized time streams:

Signal Processing and Algorithms

Preprocessing and Noise Reduction

Due to the low amplitude of BCG signals (especially from a small, ear-worn sensor), preprocessing was critical. Our pipeline included:

Beat Detection Methods

Detecting the cardiac-induced BCG fiducial points was challenging. We experimented with several methods:

Alignment and Validation

Each candidate BCG beat was aligned with the corresponding ICG-detected event. Our validation metric was the beat detection rate—the proportion of ICG beats that had matching BCG detection within an acceptable time window. This metric was computed across all experimental conditions.

The Figure in below shows example of BCG signals from the same subject during REST, INT and PLAT sessions. Matched ICG c-peaks and BCG j-peaks are shown as blue triangles and red dots, respectively. Misdetected j-peaks without matching ICG c-peaks are shown in gray.

Figure 2. Example of BCG signals from the same subject during REST, INT and PLAT sessions. Matched ICG c-peaks and BCG j-peaks are shown as blue triangles and red dots, respectively. Misdetected j-peaks without matching ICG c-peaks are shown in gray.

Experimental Results

Our analysis revealed important insights into the performance of our standalone TWS/OWS earbud approach.

Beat Detection Performance

Table 1. Summary of average BCG beat selection performances over all subjects. Best precision and accuracy (ACC) values are shown in bold. Baseline implies selecting all beats detected by the Brüser’s algorithm, resulting to 100% recall.

Conclusion

Our research demonstrates that a fully standalone TWS/OWS earbud can capture BCG signals with sufficient reliability for PTT-based BP monitoring—especially under low-motion conditions. Although motion artifacts remain a challenge, our signal processing and filtering approaches (notably IBI outlier removal) significantly improve beat dete4ction rates. With further refinements in sensor technology and algorithm design, such earbuds hold promise for seamless, cuffless BP monitoring that integrates into everyday life.

For more details, please read our full paper:
Evaluation of Wearable Head BCG for PTT Measurement in Blood Pressure Monitoring