Synheart Focus Overview

​

What is Synheart Focus?

• 🎯 Focused: Optimal cognitive state, high attention and productivity (Score: 70-100)
• 😴 Bored: Low engagement, reduced attention (Score: 30-50)
• 😰 Anxious: Heightened arousal, reduced efficiency (Score: 20-40)
• 🔥 Overload: Cognitive overload, information processing difficulty (Score: 0-20)

​

Key Features

​

On-Device Processing

• All inference happens locally on your device
• No network calls required
• No raw biometric data leaves the device
• Privacy-first by design

​

Real-Time Performance

• < 10ms inference latency per inference (ONNX models)
• >2 MB model size (Gradient Boosting ONNX)
• 60-second sliding window with 5-second steps

​

Research-Based

• Trained on SWELL dataset (stress and workload detection)
• 4-class Gradient Boosting classifier
• 24 HRV features (time-domain, frequency-domain, statistical)
• ONNX format optimized for on-device inference
• Subject-specific z-score normalization

​

Multi-Platform

​

Architecture

Wearable / Sensor
   └─(HR bpm, RR ms)──► Your App
                           │
                           ▼
                   Synheart Focus SDK
            [Ring Buffer] → [HR→IBI] → [Z-Score Normalize]
                                     │
                              [24 HRV Features]
                                     │
                                  [Model]
                                     │
                              FocusResult

• Ring Buffer: Holds last 60s of HR/RR data (configurable, default: 60s window, 5s step)
• HR→IBI Converter: Converts heart rate (BPM) to inter-beat intervals (ms)
• Z-Score Normalizer: Subject-specific normalization using adaptive baseline
• Feature Extractor: Computes 24 HRV features (time-domain, frequency-domain, statistical)
• Model: Gradient Boosting classifier (4-class: Focused, Bored, Anxious, Overload)
• Scorer: Maps probabilities to focus scores (0-100)

​

Quick Start Examples

import 'package:synheart_focus/synheart_focus.dart';

// Initialize the focus engine
final engine = FocusEngine(
  config: const FocusConfig(
    windowSeconds: 60,
    stepSeconds: 5,
    minRrCount: 30,
  ),
);

// Initialize with ONNX model
await engine.initialize(
  modelPath: 'assets/models/Gradient_Boosting.onnx',
  backend: 'onnx',
);

// Subscribe to focus updates
engine.onUpdate.listen((result) {
  print('Focus Score: ${result.focusScore}');
  print('Focus State: ${result.focusState}');
});

// Provide HR data (BPM)
await engine.inferFromHrData(
  hrBpm: 72.0,
  timestamp: DateTime.now(),
);

​

Use Cases

​

Productivity Apps

final focusEngine = FocusEngine(
  config: const FocusConfig(),
);
await focusEngine.initialize(
  modelPath: 'assets/models/Gradient_Boosting.onnx',
);

// Stream from wearable
synheartWear.streamHR(interval: const Duration(seconds: 1))
  .listen((metrics) {
    final hr = metrics.getMetric(MetricType.hr);
    if (hr != null) {
      focusEngine.inferFromHrData(
        hrBpm: hr.toDouble(),
        timestamp: DateTime.now(),
      ).then((result) {
        if (result != null && result.focusScore < 50) {
          sendNotification("Consider taking a break");
        }
      });
    }
  });

​

Wellness Coaching

// In your Flutter wellness app
final focusEngine = FocusEngine(
  config: const FocusConfig(
    windowSeconds: 60,
    stepSeconds: 5,
  ),
);

// Integrate with Synheart Wear
synheartWear.streamHR(interval: const Duration(seconds: 1))
  .listen((metrics) {
    final hr = metrics.getMetric(MetricType.hr);
    if (hr != null) {
      focusEngine.inferFromHrData(
        hrBpm: hr.toDouble(),
        timestamp: DateTime.now(),
      ).then((result) {
        if (result != null) {
          updateDashboard(result);
        }
      });
    }
  });

​

Research Applications

// Flutter research app
final focusEngine = FocusEngine(
  config: const FocusConfig(
    windowSeconds: 60,
    stepSeconds: 5,
    minRrCount: 30,
  ),
);

// Log all results for analysis
focusEngine.onUpdate.listen((result) {
  database.insert(
    timestamp: result.timestamp,
    focusState: result.focusState,
    focusScore: result.focusScore,
    confidence: result.confidence,
    features: result.features,
  );
});

​

Model Details

• Time-domain (9): mean_rr, std_rr, min_rr, max_rr, range_rr, rmssd, sdnn, nn50, pnn50
• Frequency-domain (11): VLF, LF, HF, UHF powers, total_power, lf_hf_ratio, normalized powers
• Statistical (4): skewness, kurtosis, median_rr, iqr

1. Stream HR data (1 Hz) from wearable device
2. Buffer in 60-second sliding window
3. Convert HR (BPM) → IBI (ms)
4. Apply subject-specific z-score normalization
5. Extract 24 HRV features
6. Run ONNX model inference (Gradient Boosting)
7. Calculate focus score from 4-class probabilities
8. Return result with all features and probabilities

• Latency: < 10ms per inference (ONNX models)
• Model Size: >2 MB (Gradient Boosting ONNX)
• Accuracy: Validated on SWELL dataset

​

API Parity

​

Available SDKs

​

Privacy & Security

• On-Device Processing: All focus inference happens locally
• No Data Retention: Raw biometric data is not retained after processing
• No Network Calls: No data is sent to external servers
• Privacy-First Design: No built-in storage - you control what gets persisted
• Not a Medical Device: This library is for wellness and research purposes only

​

Resources

• GitHub: synheart-ai/synheart-focus
• Flutter Package: synheart-focus-flutter
• Issues: Report Bugs
• Discussions: Community Forum

​

Citation

@software{synheart_focus,
  title = {Synheart Focus},
  author = {Synheart AI Team},
  year = {2025},
  version = {0.0.1},
  url = {https://github.com/synheart-ai/synheart-focus}
}