Global Breath-Link Initiative

Democratizing Lung Health: A Triage-to-Treatment Ecosystem

🏠 Home 🫁 Awareness Simulator 🔬 AI Acoustic Biopsy 💊 Treatment Plan 🌈 Kids Zone
👋 Need help? Ask me here!
Breathlink Support
Hello! Welcome to the AI Acoustic Biopsy setup. How can I help you extract biomarkers today?
PROTOTYPE v0.1 — KNN CLASSIFIER

AI Acoustic Biopsy

Real cough audio feature extraction compared against a reference library derived from published COUGHVID & Coswara respiratory research datasets using K-Nearest Neighbor classification.

01
Record
02
Profile
03
Analysis
04
Results

Cough Audio Capture

// WEB AUDIO API · FFT · REAL-TIME SPECTRAL ANALYSIS
FFT — 2048 bins RMS Energy Zero Crossing Rate Spectral Centroid Spectral Bandwidth Spectral Rolloff KNN Classifier
LIVE WAVEFORM
0:00
Press the microphone and perform 3 forced coughs.
Recording auto-stops after 8 seconds.

How to Cough Correctly

1
Sit upright, mic nearby
2
Take a deep breath in
3
Cough forcefully 3 times
4
Wait for analysis

Health Risk Profile

// COMBINED WITH ACOUSTIC SCORE FOR FINAL TRIAGE

KNN Classifier Running

// COMPARING YOUR COUGH AGAINST REFERENCE LIBRARY
🫁
[INIT] Loading audio feature buffer...

Acoustic Classification Complete

// KNN TRIAGE SCORE — RESEARCH-GROUNDED
--
CALCULATING

K-Nearest Neighbor Matches

// YOUR COUGH VS COUGHVID/COSWARA REFERENCE PROFILES
Research basis: Reference feature ranges derived from published studies using the COUGHVID dataset (25,000+ recordings, Orlandic et al. 2021) and Coswara dataset. Features: spectral centroid (healthy > symptomatic), ZCR (restrictive disease = high, obstructive = low), RMS energy, and spectral bandwidth (1300–3200 Hz for coughs).

Extracted Audio Features

// RAW MEASUREMENTS FROM YOUR COUGH RECORDING
⚠️ Important: This is a research prototype pre-screening tool, not a medical diagnosis. The KNN classifier uses published acoustic feature ranges as a reference — it cannot replace clinical evaluation. Always consult a qualified healthcare provider.