0x7f3a1e000000
Delirium Detector
Sensors sampled on an ESP32 and streamed to a Raspberry Pi running a trained LSTM classifier for delirium detection.
- sensor sampling
- 100 Hz
- end-to-end detection latency
- <2 s
- classification accuracy
- >85%
Firmware
ESP32 code samples at 100 Hz, preprocesses signals in C++ within 4.5 KB of statically allocated buffers, and streams frames to the host over Wi-Fi.
Model
The trained LSTM runs on a Raspberry Pi, taking frames off the network and classifying them with under 2 s of end-to-end latency.
