Realtime Estimator Selection Guide

Overview

Choosing the right realtime estimator is crucial for your application's success. This guide helps you make an informed decision based on your specific requirements, constraints, and use cases.

Quick Decision Tree

Detailed Comparison

ME-rPPG Estimator (Default)

Best For:

• Consumer fitness and wellness applications
• Challenging lighting or motion conditions
• Applications where AI robustness is valuable
• Projects with open-source requirements
• High-end devices with good connectivity

Choose ME-rPPG if:

• ✅ Users may move slightly during measurement
• ✅ Lighting conditions vary significantly
• ✅ You can download ~10 MB of models
• ✅ Open-source licensing is required
• ✅ You want state-of-the-art AI accuracy

Avoid ME-rPPG if:

• ❌ You need instant initialization
• ❌ Bundle size is critical (< 1 MB)
• ❌ Offline-first is required immediately
• ❌ Target devices have limited memory
• ❌ Slow network connections are common

FDA Estimator

Best For:

• Medical and healthcare applications
• Regulatory compliance requirements
• Resource-constrained environments
• Embedded systems and IoT devices
• Applications requiring instant startup

Choose FDA if:

• ✅ Medical-grade accuracy is required
• ✅ You need instant initialization
• ✅ Bundle size must be minimal
• ✅ Offline capability is critical
• ✅ Proprietary IP protection is acceptable

Avoid FDA if:

• ❌ Users will be moving significantly
• ❌ Lighting conditions are uncontrolled
• ❌ You need maximum motion tolerance
• ❌ Open-source licensing is required
• ❌ AI-powered features are preferred

Feature Comparison Matrix

Feature	ME-rPPG	FDA	Winner
Accuracy (Ideal Conditions)	⭐⭐⭐⭐⭐	⭐⭐⭐⭐⭐	🤝 Tie
Accuracy (Motion)	⭐⭐⭐⭐⭐	⭐⭐⭐⭐	🤖 ME-rPPG
Accuracy (Poor Lighting)	⭐⭐⭐⭐⭐	⭐⭐⭐⭐	🤖 ME-rPPG
Initialization Speed	⭐⭐⭐ (1-2s)	⭐⭐⭐⭐⭐ (instant)	📈 FDA
Processing Speed	⭐⭐⭐⭐ (10-30ms)	⭐⭐⭐⭐ (5-10ms)	📈 FDA
Bundle Size	⭐⭐⭐ (~10 MB)	⭐⭐⭐⭐⭐ (~100 KB)	📈 FDA
Memory Usage	⭐⭐⭐ (~4 MB)	⭐⭐⭐⭐⭐ (<200 KB)	📈 FDA
Setup Complexity	⭐⭐⭐ (models required)	⭐⭐⭐⭐⭐ (zero config)	📈 FDA
Licensing	⭐⭐⭐⭐⭐ (open-source)	⭐⭐⭐ (proprietary)	🤖 ME-rPPG
Motion Tolerance	⭐⭐⭐⭐⭐	⭐⭐⭐⭐	🤖 ME-rPPG
Lighting Tolerance	⭐⭐⭐⭐⭐	⭐⭐⭐⭐	🤖 ME-rPPG
First Result	~3 seconds	~5 seconds	🤖 ME-rPPG
Optimal Accuracy	~10 seconds	~10 seconds	🤝 Tie

Use Case Recommendations

Healthcare & Medical Applications

Recommended: FDA Estimator

const camera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: RealtimeEstimatorType.Fda,
    earlyEstimation: false,  // Wait for full accuracy
    minDuration: 10,
    minConfidence: 0.7,      // High threshold
    debug: false
  }
});

Reasons:

• Medical-grade accuracy
• Regulatory compliance friendly
• Deterministic results
• No external dependencies
• Instant initialization

Fitness & Wellness Apps

Recommended: ME-rPPG Estimator

const camera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: RealtimeEstimatorType.MeRppg,
    earlyEstimation: true,
    minDuration: 3,
    minConfidence: 0.3,
    debug: false
  }
});

Reasons:

• Excellent motion tolerance (users may be active)
• Robust to varying lighting
• Fast initial results
• State-of-the-art AI accuracy
• Open-source licensing

Mobile Apps (iOS/Android)

Recommended: ME-rPPG Estimator

Reasons:

• Users often move devices
• Lighting varies (indoor/outdoor)
• Modern phones have good connectivity
• Memory is generally available
• Users expect AI-powered features

Alternative: FDA Estimator (for medical apps or limited connectivity)

Embedded Systems & IoT

Recommended: FDA Estimator

const camera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: RealtimeEstimatorType.Fda,
    earlyEstimation: true,
    minDuration: 5,
    minConfidence: 0.5,
    debug: false
  }
});

Reasons:

• Minimal memory footprint
• No model downloads required
• Instant initialization
• Works offline immediately
• Smaller bundle size

Research & Development

Recommended: Both (for comparison)

// Test both estimators
const meRppgCamera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: RealtimeEstimatorType.MeRppg,
    debug: true
  }
});

const fdaCamera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: RealtimeEstimatorType.Fda,
    debug: true
  }
});

Reasons:

• Compare AI vs signal processing approaches
• Analyze different algorithm behaviors
• Understand trade-offs empirically
• Access to signal data from both

Progressive Web Apps (PWA)

Recommended: ME-rPPG Estimator (with FDA fallback)

async function createEstimator() {
  // Try ME-rPPG first
  try {
    const camera = createVitalSignCamera({
      realtimeEstimationConfig: {
        estimatorType: RealtimeEstimatorType.MeRppg,
        debug: true
      }
    });
    
    // Wait for models to load
    await new Promise(resolve => setTimeout(resolve, 2000));
    return camera;
  } catch (error) {
    console.warn('ME-rPPG failed, falling back to FDA:', error);
    
    // Fallback to FDA
    return createVitalSignCamera({
      realtimeEstimationConfig: {
        estimatorType: RealtimeEstimatorType.Fda,
        debug: true
      }
    });
  }
}

Performance Comparison

Initialization Time

Processing Performance

Metric	ME-rPPG	FDA
Frame Processing	10-30ms	5-10ms
First Result	~3 seconds	~5 seconds
Optimal Accuracy	~10 seconds	~10 seconds
CPU Usage	Medium-High	Medium
Memory Usage	~4 MB	<200 KB

Accuracy Under Different Conditions

Condition	ME-rPPG	FDA
Ideal (Still, Good Light)	±2 BPM	±2 BPM
Slight Motion	±2-3 BPM	±3-5 BPM
Moderate Motion	±3-5 BPM	±5-10 BPM
Poor Lighting	±2-4 BPM	±4-8 BPM
Talking	±3-5 BPM	±5-10 BPM

Migration Guide

Switching from FDA to ME-rPPG

Step 1: Add Model Files

# Copy ME-rPPG models to public directory
cp -r node_modules/ts-vital-sign-camera/public/models/me-rppg public/models/

Step 2: Update Configuration

// Before (FDA)
const camera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: RealtimeEstimatorType.Fda,
    minDuration: 10,
    minConfidence: 0.6
  }
});

// After (ME-rPPG)
const camera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: RealtimeEstimatorType.MeRppg,
    minDuration: 3,          // Can be faster
    minConfidence: 0.3,      // Can be lower
    modelPath: 'models/me-rppg/model.onnx',
    statePath: 'models/me-rppg/state.json',
    welchPath: 'models/me-rppg/welch_psd.onnx',
    hrPath: 'models/me-rppg/get_hr.onnx'
  }
});

Step 3: Handle Async Initialization

// ME-rPPG requires async initialization
camera.on('ready', () => {
  console.log('ME-rPPG models loaded and ready');
  enableStartButton();
});

camera.on('error', (error) => {
  console.error('Failed to load ME-rPPG:', error);
  // Fallback to FDA or show error
});

Switching from ME-rPPG to FDA

Step 1: Update Configuration

// Before (ME-rPPG)
const camera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: RealtimeEstimatorType.MeRppg,
    minDuration: 3,
    minConfidence: 0.3,
    modelPath: 'models/me-rppg/model.onnx',
    // ... other model paths
  }
});

// After (FDA)
const camera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: RealtimeEstimatorType.Fda,
    minDuration: 10,         // Longer for stability
    minConfidence: 0.6       // Higher threshold
  }
});

Step 2: Remove Model Files (optional)

# Remove ME-rPPG models to reduce bundle size
rm -rf public/models/me-rppg

Step 3: Update User Guidance

// FDA is more sensitive to motion
const instructions = [
  "Stay completely still",        // More important for FDA
  "Don't talk during measurement", // Critical for FDA
  "Ensure good, even lighting",   // More important for FDA
  "Wait for full 10 seconds"      // FDA needs more time
];

Runtime Switching

You can switch estimators at runtime by recreating the camera:

let currentEstimator: RealtimeEstimatorType = RealtimeEstimatorType.MeRppg;
let camera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType: currentEstimator
  }
});

function switchEstimator(newType: RealtimeEstimatorType) {
  // Clean up current camera
  camera.destroy();
  
  // Create new camera with different estimator
  currentEstimator = newType;
  camera = createVitalSignCamera({
    realtimeEstimationConfig: {
      estimatorType: currentEstimator
    }
  });
  
  console.log(`Switched to ${newType} estimator`);
}

// Usage
switchEstimator(RealtimeEstimatorType.Fda);

Decision Checklist

Use this checklist to make your decision:

Choose ME-rPPG if you answer "Yes" to most:

• [ ] Users may move during measurement
• [ ] Lighting conditions are uncontrolled
• [ ] You can download ~10 MB of models
• [ ] Open-source licensing is required
• [ ] You want AI-powered features
• [ ] Target devices are modern (2020+)
• [ ] Network connectivity is good
• [ ] You prioritize accuracy in challenging conditions

Choose FDA if you answer "Yes" to most:

• [ ] Medical-grade accuracy is required
• [ ] Instant initialization is critical
• [ ] Bundle size must be minimal
• [ ] Offline capability is essential
• [ ] Users can stay still during measurement
• [ ] Lighting can be controlled
• [ ] Regulatory compliance is needed
• [ ] You prefer deterministic algorithms

Hybrid Approach

For maximum flexibility, implement both and let users or conditions decide:

function selectEstimator(conditions: {
  lighting: 'good' | 'poor';
  motion: 'still' | 'moving';
  connectivity: 'online' | 'offline';
  priority: 'speed' | 'accuracy';
}): RealtimeEstimatorType {
  // Offline or poor connectivity → FDA
  if (conditions.connectivity === 'offline') {
    return RealtimeEstimatorType.Fda;
  }
  
  // Poor lighting or motion → ME-rPPG
  if (conditions.lighting === 'poor' || conditions.motion === 'moving') {
    return RealtimeEstimatorType.MeRppg;
  }
  
  // Speed priority → FDA (instant init)
  if (conditions.priority === 'speed') {
    return RealtimeEstimatorType.Fda;
  }
  
  // Default to ME-rPPG for best overall experience
  return RealtimeEstimatorType.MeRppg;
}

// Usage
const estimatorType = selectEstimator({
  lighting: detectLightingConditions(),
  motion: detectMotionLevel(),
  connectivity: navigator.onLine ? 'online' : 'offline',
  priority: 'accuracy'
});

const camera = createVitalSignCamera({
  realtimeEstimationConfig: {
    estimatorType
  }
});

Summary

Scenario	Recommended	Reason
Medical Apps	FDA	Regulatory compliance, deterministic
Fitness Apps	ME-rPPG	Motion tolerance, AI robustness
Mobile Apps	ME-rPPG	Varying conditions, modern devices
Embedded Systems	FDA	Minimal resources, instant init
PWAs	ME-rPPG + FDA fallback	Best experience with safety net
Research	Both	Compare approaches
Offline-First	FDA	No model downloads
Open-Source Projects	ME-rPPG	Licensing requirements

Next Steps

1. Review Detailed Guides:

   • ME-rPPG Estimator
   • FDA Estimator

2. Configure Your Estimator:

   • Configuration Reference

3. Optimize Performance:

   • Performance Guide

4. Handle Edge Cases:

   • Troubleshooting Guide

References

• ME-rPPG Paper
• FDA Guidelines
• SDK Documentation