Bare metal wearable for real-time Electrocardiogram (ECG), Seismocardiogram (SCG), and respiratory rate extraction.

Core Stack: STM32U5 (Cortex-M33) | ST1VAFE6AX (6-axis IMU) | AD8232 (ECG AFE)

To guarantee precise timing, the firmware relies entirely on hardware synchronization. The IMU's data-ready external interrupt triggers a simultaneous DMA transfer for both the SPI bus (IMU) and the ADC (ECG). This locks both data streams to a 480 Hz sample rate, eliminating software polling jitter.

The main loop runs three independent DSP pipelines on-device:

• ECG (Pan-Tompkins++): Bandpass -> derivative -> squaring -> flattop smoothing -> MWI. Uses adaptive thresholding with search-back for real-time R-peak detection [1].tection [1].
• SCG (Template Matching NCC): Accel Z-axis (dorso-ventral) filtered at 7-30 Hz. Bootstraps a heartbeat template during the first 10s via a 4th-power envelope and pairwise NCC [2]. Runs a sliding normalized cross-correlation against the template for continuous AO detection. The template is double-buffered and refreshed every 60s.
• Respiratory Rate: Accel Z-axis LPF -> decimated 48:1 (to 10 Hz) -> bandpass 0.1-0.6 Hz. Fed into a 30-second autocorrelation window, outputting a new respiratory rate every 5 seconds.conds.

HRV metrics (RMSSD, SDNN, pNN50, SD1, SD2) are calculated independently for both the ECG and SCG streams over a 60 second sliding window based on Task Force standards [3,4]. ECG and SCG reports are synchronized as the ECG trigger drives both computations simultaneously. Data is streamed to a host via a USB Virtual COM Port with CRC-32 integrity.

• /hardware/HM-Rev1: First spin. See ERRATA.md (known routing/schematic flaw with the AD8232).
• /hardware/HM-Rev2: Fixed AD8232 layout, added USB-C, placed components on the bottom layer.
• /fw: STM32CubeIDE project, bare-metal C source, DSP logic, and TinyUSB stack.
• /software: Python telemetry parser and plotter.

• Hardware validation.
• Synchronized IMU + ECG DMA capture.
• ECG R-peak detection (Pan-Tompkins++).
• SCG heartbeat detection (Template Matching NCC).
• Real-time respiratory rate from chest-wall accelerometry.
• On-device HRV metrics calculation.
• USB CDC telemetry via TinyUSB.
• SCG vs ECG HRV validation.
• Improve SCG RMSSD agreement
• Multi-axis respiratory fusion (PCA + gyro).

To validate the SCG pipeline, both streams are recorded simultaneously and HRV metrics are compared using Bland-Altman analysis, Pearson correlation, and ICC(2,1). Results from a 30 minute resting session (single subject, n=57 paired reports after 4 outliers removed via 3xIQR on RMSSD):

SDNN and SD2 show excellent agreement. RMSSD underestimates ECG systematically. The SCG pipeline captures long term variability accurately but doesn't perform well with beat-to-beat variations.

The validation script reports both unfiltered and filtered results. Run python software/analyze.py ir> to produce the table and plots with data saved from the software/monitor.py script.

Compiled with STM32CubeIDE.

1. git clone https://github.com/konstantinosfragkoulis/HealthMonitor.git
2. Open STM32CubeIDE > t; File > STM32 Project Create/Import > Import STM32 Project > STM32CubeMX/STM32CubeIDE ProjectE Project.
3. Point to the /fw directory.
4. Build (Release/Debug) and flash via ST-Link.

The STM32 streams binary packets over USB CDC. Use the included Python script to plot sensor data in real time.

cd software
pip install pyserial numpy matplotlib  # You might want to use a virtual environment for this


# Auto detect port
python monitor.py



# Manually select port
python monitor.py COM3          # Windows
python monitor.py /dev/ttyACM0  # Linux

• USB stack powered by the open-source TinyUSB library (MIT License).

[1] M. N. Imtiaz and N. Khan, "Pan-Tompkins++: A Robust Approach to Detect R-peaks in ECG Signals," in Proc. IEEE Int. Conf. Bioinformatics and Biomedicine (BIBM) Workshops, 2022. [Online]. Available: https://arxiv.org/abs/2211.03171

[2] S. Parlato, J. Centracchio, D. Esposito, P. Bifulco, and E. Andreozzi, "Fully automated template matching method for ECG-free heartbeat detection in cardiomechanical signals," Physical and Engineering Sciences in Medicine, vol. 48, pp. 649-664, 2025.

[3] Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology, "Heart rate variability: standards of measurement, physiological interpretation and clinical use," Circulation, vol. 93, pp. 1043-1065, 1996.

[4] F. Shaffer and J. P. Ginsberg, "An Overview of Heart Rate Variability Metrics and Norms," Frontiers in Public Health, vol. 5, 258, 2017.