Fundamentals of Biosignal Measurement and Signal Processing

From the principles of the ECG and pulse wave to filters, spectral analysis, peak detection, and accuracy assessment

Published

October 18, 2026

Introduction

English version — machine-assisted translation (by generative AI, Anthropic Claude) of the original Japanese hands-on. 日本語版はこちら.

Heart rate variability (HRV) analysis, heart rate from wearables, SpO₂ from pulse oximeters —— all of them rest on the same flow: measure a biosignal, digitize it, process it, and extract a number. If something goes wrong anywhere along the way, the final number can easily be wrong.

This material summarizes the fundamentals of biosignal measurement and signal processing assumed in, for example, the lecture material on heart rate variability (HRV). Using the electrocardiogram (ECG) and photoplethysmography (PPG) as examples, it covers in order:

  1. Signal sources (the electrical activity of the heart, absorption of light by blood)
  2. Measurement hardware (electrodes, differential amplification, noise countermeasures)
  3. Digitization (sampling, quantization)
  4. Signal processing (filters, spectral analysis, peak detection)
  5. Accuracy assessment (Bland–Altman analysis, etc.)

Almost every chapter has a demo that runs in the browser.

WarningCaution

This material is for educational purposes only. For the design and evaluation of medical devices, refer to the relevant standards (such as the IEC 60601 series) and the latest guidelines.

Intended audience

  • Students in biomedical measurement, medical engineering, and computing, and engineers and researchers who work with wearables or biosignals
  • People in medicine or sports who want to understand device readings starting from “why does it give that value?”
  • Knowledge at the level of high-school mathematics (trigonometric and exponential functions) is assumed. The Fourier transform is introduced starting from an intuitive explanation

Learning goals

  1. Explain which electrical activity of the heart each ECG waveform corresponds to, and what a lead is
  2. Explain why ECG amplifiers need differential amplification, high CMRR, and a driven right leg
  3. Explain the principle of PPG (the Beer–Lambert law) and the ratio of ratios for SpO₂
  4. Explain the Nyquist frequency, aliasing, and quantization, and choose appropriate sampling conditions
  5. Explain the types of filters used for the ECG and their side effects (phase shift, waveform distortion)
  6. Estimate spectra with an understanding of frequency resolution, window functions, and Welch’s method
  7. Explain the meaning of each stage of the Pan–Tompkins algorithm
  8. Assess the accuracy of a new measuring device with Bland–Altman analysis

Chapters

Chapter Content Demo Approx. time
1 Cardiac electrophysiology and the ECG Computing leads 20 min
2 ECG measurement hardware Computing CMRR 20 min
3 Photoplethysmography (PPG) and SpO₂ Computing SpO₂ 20 min
4 Sampling and quantization Aliasing, quantization 20 min
5 Digital filters Filtering a noisy ECG 30 min
6 Fourier analysis and spectral estimation Resolution, windows, Welch’s method 30 min
7 Peak detection: the Pan–Tompkins algorithm Visualizing each stage and detection accuracy 30 min
8 Assessing measurement accuracy Bland–Altman analysis 20 min
9 Review questions 20 min
NoteSlide version

The lecture slides with the key points are here. Press F for full screen, S for speaker notes, and Esc for the overview.

The big picture: the measurement pipeline

flowchart LR
  Src["Signal source<br/>electrical activity of the heart /<br/>changes in blood volume"] --> Sens["Sensor<br/>electrodes / LED+PD"]
  Sens --> AFE["Analog circuitry<br/>differential amplification,<br/>anti-aliasing"]
  AFE --> ADC["A/D conversion<br/>sampling,<br/>quantization"]
  ADC --> DSP["Digital signal processing<br/>filters, spectra,<br/>peak detection"]
  DSP --> Feat["Features<br/>heart rate, HRV, SpO₂"]
  Feat --> Val["Accuracy assessment<br/>Bland–Altman"]
  Noise["Noise sources<br/>mains, EMG, motion,<br/>baseline wander"] -.-> Sens
  Noise -.-> AFE

Information lost at one stage cannot be recovered at a later stage. Keep in mind as you read that “we can just filter it later” often does not hold (Chapters 4 and 5).