Notiphones Haptic Response Module

Led the design and implementation of the Response Module for Notiphones, a Purdue senior-design project that helps headphone users remain aware of their surroundings. The system was designed to recognize a user's name and deliver a discreet haptic alert without interrupting audio playback.

I owned the subsystem schematic, component selection, embedded control logic, and integration of the vibration motor, acknowledgment button, and visual status indicators. The motor was driven by a 500 Hz PWM signal so its strength could be adjusted from the companion app while remaining noticeable and comfortable. A timed cutoff prevented indefinite operation, and either the physical acknowledgment button or an app command could stop the alert immediately.

I designed power, battery, and status indication behavior; created the response-subsystem schematic; contributed to the routed system PCB; and wrote embedded C control logic and interrupt handlers for the ESP32-S3. Testing used an oscilloscope and multimeter to verify PWM behavior, current draw, alert timing, button response, and safe vibration intensity.

During cross-subsystem integration, manual vibration requests, acknowledgment events, and vibration-strength updates traveled between the Flutter app and ESP32-S3 over BLE. Team measurements recorded 0.34-second manual vibration requests, 0.19-second acknowledgment messages, and 0.54-second strength adjustments.

The response hardware, app control, and BLE communication paths were operational in the final prototype. The personalized CNN-LSTM wake-word model could be trained and transferred, but it was not executed end-to-end on the ESP32-S3 because required TensorFlow Lite modules were still missing. The final system was therefore a functional haptic and communications prototype with remaining work in on-device ML deployment and fully integrated battery validation.