DriveAware Inquire HCI v2 — Active Cockpit Interaction Prototype (Independent Developer)
Developed and implemented multimodal AI interaction components for an active cockpit prototype focused on detecting driver drowsiness cues and translating natural language to system commands. • Built a MediaPipe Face Mesh–based yawn detection module using MAR (Mouth Aspect Ratio) scoring with an exponential decay scheme. • Designed a 6-state non-blocking FSM workflow to manage monitoring, detection, inquiry, listening, confirming, and switching. • Integrated Whisper ASR, pyttsx3 TTS, and DeepSeek API for intent classification to map user speech to cockpit actions. • Created automated tests and a real-time HUD interface to evaluate state-machine logic, MAR behavior, and NLP parsing reliability.