Group course project. This was developed as part of a university course together with a project partner. I was responsible for the majority of the hardware design, and did 100% of the firmware and filter/signal processing. My project partner contributed smaller parts of the hardware design (given that he was new to hardware design) and developed the Android app.
This project implements a complete pulse oximeter system, combining:
- Custom hardware design
- Embedded firmware development
- Signal processing (filtering and peak detection)
- A simple Android app connected via Bluetooth Low Energy (BLE)
The system measures pulse and oxygen saturation and displays the results in real time on a mobile device.
Custom-designed and built PCB with integrated pulse/oximeter sensor.
The main emphasis of this project is robust pulse detection from noisy sensor data.
- Bandpass filter designed for 0.8 – 3.5 Hz
- Covers typical human heart rate range (1.0 – 3.3 Hz) with margin
- Reduces noise and motion artifacts
- Enables reliable peak detection for pulse extraction
Filtered and processed signal:
Filter robustness: the bandpass filter reliably extracts the pulse waveform whether the raw sensor input is already clean or heavily corrupted by noise.
Clean Input High-Noise Input At first, only a highpass filter was applied to compensate for low-frequency drift. After further testing with higher-noise readings, this was extended to the full bandpass approach shown above.
Note: these plots are sampled from a ring buffer with a non-compensated offset, so the high-noise(~ index 380) and filtered image (~ index 190) show a visible cut/discontinuity where the last reading wraps around to the first.
Shows real-time pulse detection directly from the device. Recording of messurment run as displayed in "Signal Processing".
Android app displaying measured pulse and oxygen saturation via BLE.
This project demonstrates a complete embedded system with a strong focus on signal processing. The combination of tailored filtering and peak detection enables accurate pulse measurement from noisy sensor data, integrated into a fully functional hardware and software solution.