Snore Doctor
Privacy-first sleep analysis. On-device machine learning detects snoring and drives adaptive audio therapy without cloud dependency.
The Idea
Snore Doctor turns your device into a private sleep lab. It provides precise event timestamps and audio therapy to encourage position changes, all while keeping your data 100% offline.
How Detection Works
Using SoundAnalysis (CoreML) on iOS and an equivalent ML engine on Android, the app analyzes audio frames in real-time. It ignores background noise and only logs confirmed snore bouts to your local database.
Your Data, Your Insights
Raw detections are transformed into meaningful heatmaps and quality scores. All history stays on-device, with optional encrypted exports available for clinical review.
Data & Workflow
How raw audio becomes actionable sleep insights.
1. Audio Capture
Continuous low-latency buffer processing. 0% disk write at this stage.
2. On-Device ML
Classification engine filters noise. Only confirmed events proceed.
3. Local Accumulation
Events, confidence scores, and therapy stats are written to Core Data / SQL.
Real-time Feed
Bout Heatmaps
Long-term Trends
Effectiveness Stats
Technology stack
Fully local audio analytics — no cloud, no account. Real-time classification, adaptive therapy, and cross-platform visualization on iOS and Android.
Core platform layers
AVAudioEngine & AudioRecord · Continuous low-latency buffering.
IntelligenceSoundAnalysis & CoreML · Real-time on-device classification.
PersistenceCore Data & SQL · Local encrypted session storage.
Real-time processing
Continuous acoustic feature extraction and confidence scoring.
Event FilteringHeuristic grouping of detections into filtered snore bouts.
Active TherapyAdaptive audio cues triggered by detection thresholds.
Data Architecture
Aggregated metadata, quality scores, and audio file references.
Raw ML classifications with confidence scores and timestamps.
Weighted effectiveness scoring of audio cues to prevent habituation.