Automated Detection of Human Emotion Jennifer Lee Quarter 1
Goal To be able to identify emotions using a low quality camera (webcam). Applications Human-Computer Interaction Alternate Reality Product Testing
Past Research Very good results Two visual-based processes About 80-90% accuracy Generally have access to high quality cameras Two visual-based processes Marker based Shadow based Anger Sadness Happiness Neutral 0.84 0.08 0.00 0.90 0.10 0.98 0.02 0.14 Analysis of Emotion Recognition using Facial Expressions, Speech and Multimodal Information (2004) http://graphics.usc.edu/cgit/pdf/papers/ICMI2004-emotionrecog_upload.pdf
Development Python (OpenCV, PIL) Read each image GUI Head Shift Adjustments Analyze Tracking Image Lighting Adjustments Webcam Identify Markers Feature Isolation Produce Tracking Image
Progress Python webcam integration Tracking Identification GUI Basic marker identification Basic marker tracking Head movement compensation Detailed marker tracking Identification GUI
Current Progress
Problems Fuzzy results Marker placement Basic tracking give blob-like results. Marker placement