Heart Rate App Accuracy vs ECG (Published Studies)
Fingertip apps: mean absolute error 2 BPM and 4.5 BPM vs ECG (Coppetti 2017, 108 patients). Face-scan apps: 7.1 BPM and 8.1 BPM. Published studies, resting adults.
- Arrhythmia & ECG
Short answer: fingertip apps are close; face-scan apps are not. In the largest hospital test we found, 108 patients measured with four iPhone apps against ECG (Coppetti et al., 2017), the two apps that use a fingertip on the rear camera averaged 2 BPM (Heart Fitness) and 4.5 BPM (Instant Heart Rate) from ECG. The two that read your face averaged 7.1 and 8.1 BPM off. A clinical pulse oximeter in the same study was 2 BPM off. That is fine for a resting spot check, not for medical decisions. A camera reading is easier to place next to a cited range: normal heart rate by age, normal walking heart rate by age, and normal sleeping heart rate by age and sex. Those tables are population references for an optical reading, not a diagnosis.
Scroll horizontally if needed
| App (study) | Method | Correlation with ECG (r) | Mean absolute error vs ECG |
|---|---|---|---|
| Heart Fitness (Coppetti 2017) | Fingertip on camera | 0.96 | 2 BPM |
| Instant Heart Rate (Coppetti 2017) | Fingertip on camera | 0.83 | 4.5 BPM |
| What's My Heart Rate (Coppetti 2017) | Face video | 0.62 | 7.1 BPM |
| Cardiio (Coppetti 2017) | Face video | 0.60 | 8.1 BPM |
| Pulse oximeter, reference (Coppetti 2017) | Finger clip | 0.92 | 2 BPM |
| HRV4Training (Plews 2017) | Fingertip on camera, HRV | rMSSD vs ECG ≈ 1.00 | Differences "trivial" (29 healthy adults, rest) |
Source: Coppetti et al., Eur J Prev Cardiol 2017 (DOI: 10.1177/2047487317702044); Plews et al., Int J Sports Physiol Perform 2017 (DOI: 10.1123/ijspp.2016-0668). Patients were at rest; motion, darker skin and irregular rhythms make results worse.
How Camera-Based Heart Rate Apps Work
The principle is straightforward. You place your fingertip over your phone's rear camera lens while the flash LED illuminates the tissue. With each heartbeat, blood surges into the capillaries of your fingertip, absorbing slightly more light. Between beats, blood volume drops and more light reaches the camera sensor.
The app's algorithm extracts this pulsatile signal from the video stream, applies filtering to remove noise, and counts peaks to determine heart rate. Some apps also attempt to calculate heart rate variability (HRV) by measuring the precise timing between consecutive beats.
This is fundamentally the same principle behind clinical pulse oximetry and wrist-based wearable heart rate sensors. The difference is in the hardware. A phone camera was never designed for this purpose. It has a rolling shutter, automatic gain control that fights against the very signal you are trying to measure, and a flash LED with a broad, uncontrolled spectrum. For a deeper look at how smartphone cameras capture PPG signals, see our article on smartphone camera-based vital signs.
The Apps We Compared
Based on download popularity and the availability of published validation data, we focused on six widely used heart rate apps.
Instant Heart Rate (Azumio)
One of the oldest and most downloaded camera heart rate apps, with over 50 million installs on Google Play. It uses the rear camera and flash for contact-based PPG. The app displays a real-time plethysmographic waveform and provides a heart rate reading within about 10-15 seconds of measurement.
Heart Rate Monitor (Azumio)
A companion app from the same developer, designed with a slightly different interface but using the same core PPG engine. Some studies have tested both apps and found identical or near-identical results, suggesting they share the same signal processing pipeline.
Samsung Health (Camera Mode)
Samsung's built-in health app includes a heart rate measurement feature that uses the rear camera and flash. It is available on Samsung Galaxy devices and works similarly to standalone heart rate apps. Samsung has gradually de-emphasized this feature in favor of their Galaxy Watch wearable sensors.
Google Fit (Heart Rate Feature)
Google Fit added a camera-based heart rate measurement feature for Pixel phones in 2021. It uses the front-facing camera instead of the rear camera, detecting subtle color changes in the user's fingertip or face. The face-based mode uses remote PPG (rPPG) technology. For more on how rPPG works, see our camera-based rPPG overview.
Cardiio
Cardiio offers both contact (fingertip on rear camera) and contactless (face-based rPPG via front camera) measurement modes. The app targets health-conscious consumers and has appeared in several peer-reviewed validation studies.
HRV4Training
HRV4Training is aimed at athletes and wellness enthusiasts who want to track heart rate variability for recovery monitoring. It uses the rear camera and flash, but its primary value proposition is beat-to-beat interval accuracy for HRV calculation rather than simple heart rate. It has been validated in multiple sports science studies.
What the Published Research Shows
Several peer-reviewed studies have tested camera heart rate apps against medical-grade reference devices (typically 3-lead or 12-lead ECG, or FDA-cleared chest strap monitors). Here is what they found.
Overall Accuracy Metrics
The most commonly reported metrics in validation studies are mean absolute error (MAE), correlation coefficient (r), and Bland-Altman limits of agreement. These give different but complementary views of accuracy.
MAE tells you the average magnitude of the error in BPM. An MAE of 3 BPM means the app is, on average, 3 beats per minute off from the reference.
Correlation coefficient (r) indicates how well the app's readings track changes in the reference. An r above 0.95 suggests the app reliably follows the true heart rate up and down. An r below 0.80 indicates the app frequently disagrees with the reference on direction or magnitude.
Bland-Altman limits of agreement define the range within which 95% of the differences between the app and the reference fall. Narrow limits (say, plus or minus 5 BPM) indicate consistently close agreement. Wide limits (plus or minus 15 BPM) indicate unreliable performance.
Study Findings by App
Coppetti et al. (2017) published one of the most rigorous comparisons, testing several smartphone apps against ECG in 108 adult patients at a Swiss hospital. Their findings, published in the European Journal of Preventive Cardiology, showed that the two fingertip-contact apps correlated well with ECG (Instant Heart Rate r = 0.83, mean absolute error 4.5 BPM; Heart Fitness r = 0.96, 2 BPM), while the two non-contact face-video apps were weaker (r = 0.62 and 0.60, errors of 7.1 and 8.1 BPM) (DOI: 10.1177/2047487317702044). The contact apps' performance is solid. Not quite medical grade, but close enough for general wellness tracking.
Samsung Health's camera mode and Google Fit's camera feature were not part of that study, and we did not find independent peer-reviewed validation for either.
Across studies, apps with more sophisticated signal processing pipelines consistently outperform apps that use simple peak detection.
HRV4Training has been specifically validated for beat-to-beat interval accuracy, which is a stricter test than simple heart rate accuracy. Plews et al. (2017) compared smartphone PPG against a Polar H7 chest strap and ECG in 29 healthy subjects at rest and found almost perfect correlations with ECG for rMSSD, with differences classed as trivial (DOI: 10.1123/ijspp.2016-0668). This makes HRV4Training one of the most clinically validated consumer heart rate apps available.
Cardiio's face-based mode was the least accurate app in Coppetti et al. (r = 0.60, 8.1 BPM mean absolute error). Remote PPG from a face video has a much weaker signal than fingertip contact, and motion or changing light make it worse. Our article on lighting conditions and rPPG accuracy covers this topic in detail.
Google Fit's heart rate feature has little independent peer-reviewed validation that we could find.
What pooled studies show
A meta-analysis of 14 studies published between 2009 and 2016 found no significant difference between smartphone camera heart rate and validated reference methods: the pooled mean difference was −0.32 BPM (99% CI −1.24 to 0.60). In adults, the correlation with the reference method was 0.90 or higher in every included study, with a pooled correlation of 0.951. The authors concluded that camera apps agree with validated methods in adults at rest and in sinus rhythm, and that they had not been validated for use in children (De Ridder et al., JMIR Cardio 2018).
Instant Heart Rate has also been tested outside the hospital. In 111 people, the app's readings agreed closely with a Polar FT7 chest-strap monitor, with intraclass correlations of 0.92 (iOS) and 0.95 (Android) at rest and 0.90 and 0.94 right after exercise. Its standard error of measurement at rest was 5.77 BPM on iOS and 4.96 BPM on Android, compared with 4.56 BPM for the Polar monitor (Mitchell et al., Physiotherapy Theory and Practice 2016).
What Affects App Accuracy?
Even the best-performing apps are not equally accurate for all users in all conditions. Several factors introduce variability.
Skin Tone
This is the most significant and most concerning source of variability. Camera PPG relies on detecting light absorption changes through the skin. Melanin absorbs light across a broad spectrum, reducing the signal available for PPG measurement. Multiple studies have documented reduced accuracy in individuals with darker skin tones.
The effect is particularly pronounced in camera-based PPG because phone flash LEDs emit broadband white light rather than the specific red and infrared wavelengths optimized for PPG. Green wavelengths, which are present in white LED light and are strongly absorbed by melanin, dominate the camera PPG signal. This means skin tone affects smartphone PPG more than it affects clinical pulse oximeters using red and infrared LEDs.
To understand how clinical-grade devices differ from consumer apps in handling skin tone variability, read our dedicated comparison article.
Motion
Any movement during measurement introduces noise. Even subtle hand tremor can corrupt the signal. Most apps instruct users to hold still, and some display a signal quality indicator to flag excessive motion. Testing yourself while walking or exercising with a finger-on-camera app is essentially useless.
Lighting Conditions
For contact-based measurement (fingertip on rear camera), ambient lighting matters less because the flash provides the illumination. But some apps allow the flash to be turned off, in which case ambient light becomes the only light source and measurements become far less reliable.
For face-based (contactless) measurement, lighting is everything. Bright, stable, diffuse lighting produces the best results. Flickering fluorescent lights, direct sunlight, and dim environments all degrade accuracy.
Phone Model and Camera Hardware
Not all phone cameras are created equal. Frame rate, sensor sensitivity, automatic exposure algorithms, and flash LED brightness all vary between models. An app that performs well on a Samsung Galaxy S24 may perform differently on a budget Android device with a lower-quality camera sensor.
Some developers, like the team behind HRV4Training, specifically test and optimize for popular phone models. Others do not, leading to inconsistent cross-device performance.
Finger Pressure
Pressing too hard on the camera lens occludes the capillary bed and flattens the pulsatile signal. Pressing too lightly allows ambient light leakage and reduces signal quality. There is a sweet spot of moderate, consistent pressure. Most apps provide on-screen guidance, but achieving the right pressure is a learned skill.
Heart Rate Range
Accuracy tends to be best at resting heart rates (60-100 BPM). At very low rates (below 50 BPM) or elevated rates (above 120 BPM), some apps show increased error. This is partly algorithmic. Many apps use signal processing parameters tuned for the typical resting range, and performance at the extremes simply was not optimized.
Children and fast heart rates. In a study of 40 children (median age 4.3 years) using four free apps checked against an ECG monitor, accuracy was similar at the finger or toe and at the earlobe when heart rate was below 120 BPM. At 120 BPM and above, the finger/toe accuracy rate fell to 27% versus 65% at the earlobe, and the authors advised against using the apps routinely for medical purposes (Ho et al., Pediatric Cardiology 2014).
Accuracy Comparison Summary
Only one study compared several heart rate apps against ECG in the same adult patients, so this ranking rests mostly on it. A 14-study meta-analysis supports the wider finding that fingertip camera apps agree with reference methods in resting adults:
- Fingertip contact apps (Heart Fitness, Instant Heart Rate): 2 to 4.5 BPM mean absolute error vs ECG at rest (Coppetti 2017). HRV4Training's fingertip method also matched ECG for resting rMSSD (Plews 2017).
- Face-scan apps (What's My Heart Rate, Cardiio): 7.1 to 8.1 BPM mean absolute error (Coppetti 2017).
- Samsung Health and Google Fit camera modes: no independent peer-reviewed comparison found, so we can't rank them.
These numbers come from resting adults. Expect worse results while moving, in poor light, or with an irregular rhythm.
Clinical-Grade vs. Consumer-Grade: Where to Draw the Line
It is tempting to view a heart rate app with an MAE of 2-3 BPM as "almost clinical grade." But the distinction between clinical and consumer devices is not just about average accuracy.
Clinical-grade devices must perform reliably across the full range of patients: dark and light skin, neonates and elderly, healthy and critically ill, moving and still. They must meet specific FDA or CE regulatory standards with documented evidence. They must handle arrhythmias, low perfusion, and artifact conditions gracefully. And they must alert clinicians to unreliable readings.
Camera heart rate apps are validated in narrow conditions, on limited populations, and they carry no regulatory clearance for medical use. An MAE of 3 BPM in a study of 50 healthy college students does not tell you much about how the app performs in a 75-year-old patient with atrial fibrillation and cold hands.
For more on what separates clinical and consumer PPG accuracy, see our article on PPG heart rate accuracy.
Practical Recommendations
If you want to use a camera heart rate app, here are some evidence-based tips:
-
Choose an app with published validation data. HRV4Training and Instant Heart Rate both have peer-reviewed studies supporting their accuracy. Apps without published validation are a gamble.
-
Use contact mode (finger on rear camera) rather than face-based contactless mode. Contact mode is consistently more accurate in every study that has tested both approaches.
-
Measure at rest, while sitting. Hold still for the entire measurement duration. Do not talk or move your hand.
-
Apply moderate, steady finger pressure. Too much compresses the blood vessels; too little lets in ambient light.
-
Measure in a well-lit room if using the camera flash, or ensure the flash is enabled if you are in a dim environment.
-
Do not rely on these apps for medical decisions. They are reasonable for general wellness tracking and fitness logging. They are not substitutes for clinical heart rate monitors, particularly if you have a known cardiac condition.
-
Take multiple readings and average them. A single reading can be off by 5-10 BPM due to random variation. Three readings taken 30 seconds apart will give you a more reliable estimate.
The Role of Signal Processing
Why do some apps outperform others when they all use the same camera hardware? The answer is signal processing. The raw video stream from a phone camera is messy. It contains respiratory modulation, motion artifact, ambient light fluctuations, and automatic gain control adjustments that the camera firmware applies without your permission.
Better apps use advanced filtering techniques: bandpass filtering tuned to cardiac frequencies (0.7-3.5 Hz), adaptive noise cancellation, and sophisticated peak detection algorithms that can handle irregular intervals. The best apps, like HRV4Training, essentially run a miniature version of clinical-grade PPG processing on your phone.
For more on what goes into extracting a clean PPG signal, see our primer on PPG signal basics.
- Apple Watch Heart Rate: Clinical Evidence Review — wrist-based optical PPG validation
- How Accurate Is Oura Ring? — finger-based PPG accuracy comparison
- How Accurate Is WHOOP? — wrist PPG for recovery tracking
- Garmin vs Apple Watch Comparison — wrist PPG sensor comparison
- Apple Watch vs Oura Ring Comparison — wrist vs finger measurement site analysis
- Normal Heart Rate by Age and Sex: NHANES seated-pulse percentiles from infancy to 80+, plus adult medians by sex
- Normal Walking Heart Rate by Age: brisk-walk band by age, from published maximum-heart-rate equations
- Normal Sleeping Heart Rate by Age and Sex: overnight ring PPG averages by age and sex
- Resting Heart Rate Chart by Age and Sex: percentiles from NHANES survey measurements
- VO2 Max Chart by Age and Sex: percentiles from FRIEND registry treadmill tests
- What Is a Good HRV Score?: why HRV needs a personal baseline and recording context
Frequently Asked Questions
- Are heart rate apps on phones accurate?
- Fingertip camera apps can be close. In a 108-patient hospital study (Coppetti et al., 2017), Heart Fitness averaged 2 BPM from ECG and Instant Heart Rate 4.5 BPM, while two face-scan apps averaged 7.1 and 8.1 BPM. Results depend on the app, phone, skin tone, motion and finger placement. These are wellness readings, not a diagnosis.
- Which heart rate app is the most accurate?
- In the one head-to-head hospital study we found (Coppetti et al., 2017), Heart Fitness was the most accurate of four apps (r = 0.96, mean absolute error 2 BPM), followed by Instant Heart Rate (r = 0.83, 4.5 BPM). For HRV rather than heart rate, HRV4Training's camera method matched ECG rMSSD with near-perfect correlation in 29 healthy adults at rest (Plews et al., 2017). All of these use fingertip contact on the rear camera.
- Can my phone measure heart rate without a smartwatch?
- Yes. A fingertip over the rear camera and flash is how the contact apps in Coppetti et al. (2017) were tested. In that 108-patient hospital study, those apps averaged 2 BPM and 4.5 BPM from ECG. Face-scan apps in the same study averaged 7.1 BPM and 8.1 BPM. These are resting readings, not a medical test.
- Do heart rate apps work for people with darker skin tones?
- Camera heart rate apps show reduced accuracy for users with darker skin pigmentation. The broadband white LED flash and green-dominant camera signal are more affected by melanin absorption than the red and infrared wavelengths used in clinical pulse oximeters. If you have darker skin and want reliable readings, apply firm (but not excessive) pressure to ensure good optical contact, use the rear camera with flash enabled, and consider validating your app against a reference device before relying on it.
- Can heart rate apps detect atrial fibrillation?
- Some apps claim to detect irregular heart rhythms, including atrial fibrillation (AFib). The PPG signal does contain information about rhythm regularity, and research has shown that algorithms can detect AFib from PPG with reasonable sensitivity and specificity. However, detecting AFib from a 30-second fingertip recording is much harder than detecting it from a continuous wrist-based PPG signal or ECG. If you suspect you have an irregular heartbeat, see a doctor rather than relying on an app.
- Why does my heart rate app give different readings each time?
- Variation between measurements is normal and comes from several sources. Your actual heart rate fluctuates continuously in response to breathing, posture, stress, and other factors. Additionally, slight differences in finger placement, pressure, and motion introduce measurement noise. If readings vary by 3-5 BPM between attempts, that is within normal range. If readings vary by more than 10 BPM, something is wrong with your technique: check finger placement, hold still, and ensure the flash is on.
- Is a chest strap more accurate than a heart rate app?
- Usually, yes. Electrode chest straps such as the Polar H10 measure the electrical R-R interval directly through skin electrodes, which is fundamentally more reliable than optical PPG measurement through a phone camera. If accuracy matters for your use case, whether athletic training, HRV monitoring, or health tracking, a chest strap remains the gold standard among consumer devices.
- How accurate is the Instant Heart Rate app?
- In a 108-patient hospital study, Instant Heart Rate had a mean absolute error of 4.5 BPM against ECG, with a correlation of 0.83 (Coppetti et al., 2017). In 111 adults compared with a Polar chest-strap monitor, its agreement was high at rest and right after exercise, with intraclass correlations of 0.90 to 0.95 on iOS and Android (Mitchell et al., 2016). It works best as a resting spot check, not as a medical test.
Related articles
Fitbit Heart Rate Accuracy: What Studies Show
How accurate is Fitbit heart rate? Validation studies across Charge 6, Sense 2, Versa 4 and Inspire 3, and how skin tone, fit and exercise type change it.
How Accurate Is Oura Ring? Sensor Data vs Clinical Studies
A detailed analysis of Oura Ring accuracy across heart rate, HRV, sleep staging, SpO2, and temperature metrics, with clinical study data from de Zambotti et al. (2019), Kinnunen et al. (2020), and peer-reviewed validation research comparing finger-based PPG to gold-standard measurements.
PPG vs ECG Accuracy Comparison: Heart Rate, HRV, and Arrhythmia Detection Data
PPG vs ECG accuracy compared for heart rate, HRV metrics, arrhythmia detection, and respiratory rate with published validation data and comparison tables.
PPG Heart Rate Measurement: Algorithms, Accuracy, and Practical Implementation
How PPG measures heart rate step by step. Learn peak detection algorithms, beat-to-beat vs averaged HR, and accuracy benchmarks for rest, exercise, and sleep.
How Accurate Is Garmin VO2 Max? Data, Studies & Limitations
An in-depth analysis of Garmin VO2 max accuracy examining the Firstbeat algorithm, independent validation studies such as Passler et al. (2019), factors affecting estimates, and how to interpret your Garmin fitness age and VO2 max readings against laboratory-measured values.
Devices in this article
Research records with the datasets, papers and published validation results for each device.