arrow_back Back to Analyzer

📊 Accuracy & Performance Guide

Understanding Confidence Scores and Measurement Accuracy

watch Wearable vs Camera Accuracy

While camera-based rPPG provides convenient spot-checks, wearable devices offer superior accuracy and continuous monitoring. Medical-grade wearables connected via dedicated APIs provide the most reliable continuous health data.

Feature Camera (rPPG) Wearable Devices
Accuracy ±5 BPM typical ±1 BPM medical-grade
Monitoring Single point 24/7 continuous
Biomarkers 5-10 basic 50+ comprehensive
Environmental Needs good lighting Works anywhere
Movement Must stay still During any activity
Sleep Tracking Not possible Full sleep stages
Explore More Qalarc Projects →

analytics Camera-Based Accuracy Benchmarks

Metric Optimal Conditions Good Conditions Poor Conditions Medical Device
Heart Rate ±2 BPM ±5 BPM ±10 BPM ±1 BPM
HRV (RMSSD) ±5 ms ±10 ms ±20 ms ±2 ms
Respiratory Rate ±2 BPM ±3 BPM ±5 BPM ±1 BPM
Age Estimation ±3 years ±5 years ±10 years N/A
Gender Detection 95%+ 85-95% <85% N/A
Expression 85%+ 70-85% <70% N/A
Ethnicity 60-75% 45-60% <45% N/A

speed Understanding Confidence Scores

85-100% - Excellent

High accuracy, medical-grade approximation. Results are highly reliable for wellness monitoring.

70-85% - Good

Acceptable accuracy for most wellness applications. Minor variations expected.

50-70% - Fair

Moderate accuracy. Use for trends rather than absolute values. Improve conditions if possible.

<50% - Poor

Low reliability. Check environmental conditions and follow optimization tips.

wb_sunny Factors Affecting Accuracy

🔆 Lighting Conditions

  • Natural daylight or bright white LED
  • Even illumination across face
  • Avoid colored lights
  • Avoid backlighting
  • Avoid darkness or shadows

📏 Distance & Position

  • 30-60cm from camera
  • Face centered in frame
  • Full face visible
  • Avoid extreme angles
  • Don't cover face regions

🎯 Movement & Stability

  • Remain still during measurement
  • Stable camera/device position
  • Natural breathing
  • Minimize talking
  • Avoid rapid movements

👤 Personal Factors

  • Clean face (no heavy makeup)
  • Remove glasses if possible
  • Adequate skin exposure
  • Facial hair may reduce accuracy
  • Face masks block detection

💻 Technical Factors

  • HD camera (720p minimum)
  • 30 FPS capture rate
  • Stable internet connection
  • CPU usage affects processing
  • Low resolution reduces accuracy

⏱️ Duration & Sampling

  • Minimum 15 seconds recording
  • 30-60 seconds optimal
  • Continuous mode for trends
  • Short samples less accurate
  • Interrupted sampling fails

psychology Demographics Accuracy Details

Age Estimation (±3 years)

The ±3 years range represents the typical margin of error. Factors affecting age estimation:

  • Lighting can add/subtract perceived years
  • Expressions affect wrinkle visibility
  • Skin condition and health status
  • Individual aging patterns vary widely

Ethnicity Classification

Our system recognizes broad geographic ancestry patterns:

  • East Asian - China, Japan, Korea, etc.
  • European - Western & Eastern Europe
  • African - Sub-Saharan Africa
  • South Asian - India, Pakistan, Bangladesh
  • Middle Eastern - Arab, Persian regions
  • Latin American - Central & South America
  • Mixed/Other - Multi-ethnic or unclear

⚠️ Note: Human diversity is complex. These are statistical approximations only.

tips_and_updates Pro Tips for Best Results

  1. Perform measurements at the same time of day for consistency
  2. Ensure your face fills 30-40% of the camera frame
  3. Use the highest quality camera setting available
  4. Close other applications to free up processing power
  5. Take multiple measurements and average the results
  6. Record for at least 30 seconds for HRV accuracy
  7. Breathe normally - don't hold your breath
  8. Remove glasses if they cause reflections
  9. Ensure ROI quality indicators show >80%
  10. Check SNR value - should be >10 dB for good results

show_chart Interpreting Signal Quality Indicators

Indicator Excellent Good Fair Poor
SNR (Signal-to-Noise) >15 dB 10-15 dB 5-10 dB <5 dB
Motion Score 90-100% 70-90% 50-70% <50%
Lighting Score 80-100% 60-80% 40-60% <40%
ROI Coverage 90-100% 75-90% 60-75% <60%
Face Stability 95-100% 85-95% 70-85% <70%

warning Important Limitations

rocket_launch Ready for Production-Grade Biomarkers?

Camera detection is great for demos and prototypes. For real applications, you need real data.

🏥 Healthcare Apps

FDA-cleared device integration for telehealth and RPM

🏃 Fitness Platforms

Real-time activity and recovery metrics from any wearable

🧠 Mental Health

Stress, sleep, and mood insights from physiological data

🔬 Research Studies

Validated biomarkers for clinical trials and studies

Built by Qalarc — open source health tech

🚀 View All Projects 💻 GitHub