This paper discusses facial expression recognition based on the Eigen faces method and using depth data from a ToF (Time-of-flight camera) camera in order to resist varying lighting conditions in real-time applications and computer games. For each frame, the system extracts a matrix transformation for a bitmap image, and applying the same matrix to a depth image counterpart. In the training state, the system uses a Support Vector Machine and performs Principle Component Analysis directly on the depth data rather than bitmap data. As a result, the system can recognize facial expressions in extremely varying lighting conditions.
The image is from Pixar.com