Classifies sleep stages — REM, NREM 1/2, and deep — using motion density and breathing rate patterns captured through WiFi CSI on the ESP32-S3. No camera, no wearable; signals are drawn from ambient radio scattering.
Real-time biosignals derived from WiFi CSI scatter. Motion density, breathing rate, and per-subcarrier amplitude update at 10 Hz.
Full-night sleep architecture. Each step represents a 30-second epoch classified by the model. Hover to inspect cycles.
Stream of classifier emissions: sleep_stage, rem_detected, deep_sleep, sleep_quality. Streaming from the cog over MCP.
Past nights — sleep architecture and quality trends across the last 14 nights.
ESP32-S3 sensor module capturing WiFi Channel State Information. CSI is fed through breathing extraction and motion-density estimators before classification.
name: dream-stage version: 0.9.0 category: health difficulty: hard size: 26 KB hardware: - ESP32-S3 inputs: - csi.motion - csi.breathing events: - sleep_stage - rem_detected - deep_sleep - sleep_quality source: cognitum-one/cogs/src/cogs/dream-stage
Tune classifier behaviour, event emissions, and capture parameters.