Dr. Vijayan Asari, DirectorPhone: 937-229-4504
3-D Expression Recognition
Automatic 3D Facial Expression Recognition
The objective of our research is to determine Automatic 3D facial Expression Recognition using live video sequences captured by a camera, this process of exploring human emotions is one of the active research areas of Signature Science Exploration (SSE) Lab.
Several Facial Behavior Identification states are mentioned below:
- Emotive states (surprise, fear joy, sadness, disgust, anger, contempt)
- Emblematic states (blink, head nod, eye roll, shoulder shrug)
- Concealment states and deceptive states (subtle expression, micro expressions, lying)
- Cognitive states (Concentration, frustration, fatigue)
Applications: Human intelligence, interrogation, resolving autism spectrum disorders and more.
The Expression Recognition process is described below:
Head Acquisition: Acquiring real time face data (2D/3D)
Feature Extraction: Identify modes of the data that directly pertain to expression changes of human face
Expression Classification: Mapping the features to discrete expressions
We are developing an Expression Recognition system using a Photon-X Camera System that can be used for Image Acquisition. Humintell provides rules defining the mapping between Action Units (AUs) and an Expression. The system as a whole would provide translation between raw face data to human emotion. We have classified basic head nod motions and are in the process of classifying Emotive states.
The process consists of Feature Extraction that involves using Active Shape Model (ASM) to capture the landmark points, Feature tracking module tracks the landmark points from frame to frame basis using Lucas-Kanade tracker. Then using statistical pattern recognition techniques, we identify the expressions.