Facial Expression Recognition using Wavelet-Based Salient Points and Subspace Analysis Methods

There are different solutions to facial expression recognition problem. In this paper a novel method for recognizing six basic facial expressions plus neutral state is developed. The solution is based on principal component analysis of two basic parts of the face: eyes and mouth. The recognition rate of the proposed method highly depends on how accurate basic facial components are extracted. A new method based on wavelet-based salient points is used for this aim resulting in a high facial expression recognition rate as well as low computational complexity. An emotion subspace is defined for each kind of facial components that facial components are projected onto. Finally, recognition would be done using a soft recognition algorithm. Simulation results cover two different cases. In the first case, the same set of individuals is used for both training and testing. However, some new individuals are added to the test set in the second situation. It is shown that this system achieves better results compared to most of similar techniques