Real-Time Emotion Recognition System
Webcam-based affective computing and landmark-driven decision-support prototyping
This work investigated an end-to-end real-time emotion-recognition system using webcam video, landmark extraction, and machine-learning classifiers. A prototype connected inferred emotional signals to a digital-twin-oriented decision-support concept.
Research period: March 2021 to February 2022
Funder: Small and Medium Business Technology Information Promotion Agency, South Korea
Role: Project Developer
Research objectives
- Explore real-time recognition of visible emotional expressions from video.
- Examine how emotion-recognition outputs might be incorporated into a personalized digital-twin workflow.
Contributions
- Collected a custom webcam dataset with varied expressions and capture conditions.
- Used MediaPipe for preprocessing and face, hand, and body landmark extraction.
- Trained and compared machine-learning models for emotion recognition.
- Built an end-to-end prototype for linking inferred expressions to a digital-twin decision-support interface.
Methods
Computer vision, MediaPipe, facial and body landmarks, video processing, machine-learning benchmarking, affective computing, digital-twin prototypes.
This was a research prototype, not a clinically validated diagnostic system.