20 K. Srinivasa Reddy International Journal of Computer & Mathematical Sciences IJCMS ISSN 2347 – 8527 Volume 7, Issue 3 March 2018 A New Approach for Facial Expression Recognition using Non Uniform Local Binary Patterns K. Srinivasa Reddy Department of Computer Science and Engineering, Institute of Aeronautical Engineering, Hyderabad, India. ABSTRACT Facial expression recognition has attracted more and more attention due to its important applications in a wide range of areas like data driven animation, human machine interaction, robotics, driver fatigue detection etc. Facial expressions also play a major role in interpersonal communication and imparting intelligence to computer for identifying facial expressions is a crucial task. This paper presents an efficient approach for facial expression recognition by deriving a new set of stable transitions on Local Binary Pattern for selecting the Significant Non Uniform Local Binary Patterns. The proposed Significant Non Uniform Local Binary Patterns are stable, because it considered the transitions from two or more consecutive zeros to two or more consecutive ones. The proposed Significant Non Uniform Local Binary Patterns along with Uniform Local Binary Pattern features improved facial expression recognition rate. A distance function is used on proposed texture features for effective facial expression recognition. To eliminate most of the effects of illumination changes that are present in human facial expression an efficient preprocessing method is used that preserves the significant appearance details that are needed for facial expression recognition. The experimental analysis was done on the popular Japanese female facial expression (JAFFE) facial expression database and it has given state-of-the-art performance. KEYWORDS—facial expression recognition, local binary pattern, uniform local binary pattern, illumination, preprocessing, distance function, and stable transition. I. INTRODUCTION The face of a human being conveys a lot of information about identity and emotional state of the person. Facial expression is an immediate and effective part of communication among humans which carries crucial information about the mental, emotional and even physical state of a human being. It is a desirable feature of next generation computers, which can recognize facial expressions and responds accordingly and enables better human machine interactions, driver state monitoring, medical and aiding autistic children. Due to its wide range of applications, automatic facial expression recognition has attracted much attention in recent years [4,12,20]. Though much progress has been made [1,2,5,10,19], recognizing facial expression with a high accuracy remains difficult due to the subtlety, complexity and variability of facial expressions. However, the inherent variability of facial images caused by different factors like variations in illumination, pose, alignment, and occlusions makes expression recognition a challenging task. Major categories of the facial expressions are anger, disgust, fear, happy, sad, surprise and neutral. These expressions are differentiated by very fine variations in muscular movements and hence local feature extraction to represent expressions is a critical task. Features are extracted from the original facial expression images which minimizes the within class variation of expression and maximizes the between classes variations. If improper features are used, even the best classifier could not recognize proper expressions. There are two main types of approaches to extract facial expression features [18] geometric feature based methods [4,12] and the appearance based methods [10,20]. Geometric feature based methods extract geometric information from the facial expression images. In appearance based methods, features are either extracted from the entire facial expression or specific regions in facial expression images. Because of more effectiveness, we have chosen appearance based approach. Gabor wavelet appearance features were demonstrated to be more effective than geometric features [5]. However Gabor Wavelet representation is computationally expensive.