Realtime detection, trainable classifier from reference audio, distance estimator, and multilateration tool
Microphone capture & realtime detection
Status: Idle
Reference dataset & classifier training
Upload labeled reference audio (short WAV/MP3). The app extracts spectral features and trains a simple k-NN classifier in-browser to distinguish gunshots vs confusers.
Event log, features, and distance estimate
Each detected impulse includes spectral features, classifier result, an estimated distance (meters) based on calibrated amplitude decay, and an echo-analysis score (used to help filter confusers).
Multilateration from multiple devices' logs
Upload CSV logs (this device or others) with timestamp (ISO), latitude, longitude. The app computes estimated origin by solving for emission time and location using time differences of arrival (TDOA). Timestamps must be synchronized (or corrected) for millisecond accuracy for high precision.