Smart Helmet: Voice-to-Turn-Signal
Cycling helmet converting voice commands to LED turn signals. ESP32, TensorFlow Lite voice model, hands-free safety.
Overview
The Smart Helmet integrates a TensorFlow Lite voice recognition model into an ESP32-based wearable to convert spoken commands — 'left,' 'right,' 'stop' — into LED turn signals in real time. The system addresses the danger of hand-signal communication in high-traffic environments, providing cyclists with a hands-free, voice-activated alternative that keeps both hands on the handlebars.
Challenge
Running a voice recognition model on a microcontroller with limited compute and battery requires aggressive model optimization. The challenge was achieving sufficient accuracy on a TensorFlow Lite model small enough to run on ESP32 hardware while handling variable ambient noise from wind and traffic.
Outcomes
- On-device voice recognition via TensorFlow Lite on ESP32
- Sub-200ms command recognition-to-LED activation latency
- Three-command vocabulary: left, right, stop — with directional LED array
- Hands-free operation validated in live cycling conditions