Effective face detection, feature extraction & neural network based approaches for facial expression recognition

Face expression recognition is a typical task to make human and machine interaction possible. Besides this, medical science and other applications demand for such system. This paper focusses on importance of face detection and its feature parts. For this, Viola - Jones algorithm was implemented. The crucial part of this paper is feature extraction and the algorithm used for the purpose is modified local binary patterns algorithm. The results of feature extraction algorithms are compared to that in the literature to show the limitations of the existing algorithms and their suitabilility for this application. Neural network based approaches are used for classification of facial expressions. This paper enumerates two classification techniques for the facial expressions. The system performs well by the method used by us and gives efficient results. This was experimented using the Taiwanian database and the Japanese database.

I. Introduction

Computer vision field is the branch of artificial intelligence [1] that usually focuses on making computers to imitate human vision, including learning, making decisions and performing necessary actions based on visual inputs, i.e. Images. Computer vision also plays a major role in pattern recognition. Human computer interaction (HCI) becomes more effective if computer can predict about emotional state of a person and hence mood of a person from supplied images on the basis of facial expressions can be classified by computer. Mehrabian [2] found in his research that 55% of human communication information is conveyed by facial expressions. This shows that face expressions recognition is very important for face to face communication. For grouping facial expressions into multiple categories, it is necessary to extract facial features which will helps in correct identification of proper and particular expressions. Face detection from entire image, feature extraction and classifier development are the three basic but typical steps for efficient classification of facial expressions. This paper explains about the flow for facial expression detection and way to avoid illumination problems. Viola - Jones algorithm is discussed with its importance for face detection and face parts. Rule based classification [3]; PCA [4] with fuzzy [5] C-means clustering and feed forward back propagation neural network is used as a classifier for classifying the expressions of supplied face into four basic categories like neutral, surprise, sad and happy. For face portion segmentation basic image processing operation like projection and skin colour polynomial model [6] are used.