C. Fan, F. Dadgostar, A. Sarrafzadeh, H. Gholamhosseini, and M. Johnson (New Zealand)
Facial Expression Recognition, Support Vector Machine, Contour Detection, Polygon Approximation Algorithm. 1 .
Automatic facial expression recognition is one of the interesting topics in computer vision and affective computing. Using the traditional methods for training may take hundreds of hours of a typical personal computer. Therefore reducing the size of the training data, together with keeping its quality, will dramatically decrease the training time without losing recognition accuracy. In this article we have introduced an algorithm of vectorizing some components of the face that are important in facial expression recognition. We have employed polygon approximation in our algorithm to extract facial features and apply to the training of a Support Vector Machine (SVM). By using this algorithm the size of the training data will be approximately 50 times less than the original images, that is a significant improvement for training of a SVM-based classifier.
Important Links:
Go Back
