The entry point is one live waveform. The ceiling is a defensible experiment—or a published-method replication—in physiology, instrumentation, and signal processing.
For high-school and college researchers, instructors, independent developers, and anyone ready to build evidence from raw PPG.
Null results, rejected windows, and failed replications are evidence. What matters is whether your method, controls, raw data, and uncertainty are clear enough for someone else to inspect.
A PulseSensor kit, a compatible board with an analog input, USB cable, WebSerial or Arduino with PulseSensor Playground, and a place to save raw data and trial notes.
Instrument checkout · about one minute
0. Read a live PPG waveform.
Connect the sensor, open WebSerial, and rest a finger on it. Before forming a hypothesis, learn what stable signal, drift, and artifact look like in real time.
Look for a notch or secondary wave on the falling edge of a beat. In PPG research, that morphology is discussed in relation to aortic valve closure and reflected arterial waves—not just pulse counting.
Your first task is instrument literacy: observe before you measure.
Observe the waveform, warm your hand gently under water, dry it completely, and observe again. Note what changes without treating one trace as a diagnosis.
Motion as interference
Compare a still hand with a few seconds of gentle finger movement. Learn how quickly motion can dominate the physiological waveform.
Noticed something you cannot explain? Good. The studies below show how to turn that observation into a controlled question.
Research ladder · three entry studies
Begin with a result you can defend.
These are not toy exercises. Each isolates a real measurement problem: repeatability, artifact, or sampling. Start where your method is credible; advance by adding controls, trials, reference signals, or participants.
Foundation1 sensor30–60 minutes
1. Establish a repeatable PPG recording
Question: Which waveform features remain stable beat to beat, and what rule separates usable signal from rejected data?
AcquireRecord one trace under documented conditions and preserve the raw waveform.
Keep the sameSame finger, same posture, same attachment, same gentle pressure, same lighting.
MeasureMark peaks, baseline drift, repeated morphology, and any interval that fails your usability rule.
ReportShow accepted and rejected sections together, with the same decision rule applied to both.
Deliverable: one labeled waveform, acquisition notes, and a reproducible rule for accepting or rejecting signal.
For work that will be judged, taught, extended, or compared with literature, run repeated trials, save the raw waveform rather than only BPM or screenshots, label every file, and define rejection rules before seeing the result.
Question State one falsifiable question.
Baseline Define the comparison condition.
Perturbation Change one variable.
Repeat Save raw data and metadata.
Report Show uncertainty, exclusions, and failure.
Write the study plan
Research question: ___ Expected direction: ___ Independent variable: ___ Outcome measures: ___ Controlled variables: ___ Trials and rejection rules: ___ Validation or reference signal: ___
Advanced · published-method replication
4. Rebuild a published PulseSensor study.
Replication is a demanding systems test: hardware, timing, signal processing, physiology, and reporting must all line up closely enough for another group’s result to appear in your data.
Question: Can four PulseSensors recover heart rate and respiratory rate the way the published method says they should?
Published benchmark: Stevens and colleagues (2024) used four PulseSensors to collect PPG from 15 participants, derived heart and respiratory rates with off-the-shelf hardware and open-source software, and compared the results with commercial wearables. The method is documented; your job is to test what reproduces.
ReconstructFollow the published method as closely as your setup allows and document every departure.
Hold constantPreserve sensor sites, sampling rate, recording length, posture, and processing choices across trials.
ValidateTrack agreement, drift, channel failures, rejected windows, and performance against an appropriate reference.
ReportPlace your result beside the published result and explain differences rather than hiding them.
Deliverable: a replication record containing the protocol, raw data, deviations, analysis, and failures. A well-documented failure to reproduce is a real result.
Before you collect participant data
Participant protections are part of the protocol. Before recruiting anyone or collecting data for a student or institutional study—including some self-testing—ask your teacher, adult sponsor, or institution which review applies. Obtain any required IRB approval and consent or assent before data collection, and do not collect or publish identifiable health data.
Then choose another entry on PulseSensor in Published Research. Each paper contributes a method, dataset, constraint, or benchmark that can become the next study.
Know what this sensor measures.
PulseSensor records one green-light reflectance PPG channel. It does not measure SpO₂ or blood pressure and is not medical equipment. Breathing, timing, and variability observations remain experimental until validated against an appropriate reference.
