Expression Recognition

www.ee.surrey.ac.uk/Personal/R.Bowden ABSTRACT: This manuscript and associated talk gives an historical but not exhaustive overview of work in the Cognitive Vision Lab at the University of Surrey’s Centre for Vision Speech and Signal Processing. Work concentrates on people, tracking or identifying their actions and interpreting the meaning of those actions. To do this we employ techniques from a variety of sources which include the use of Mutual Information in Tracking, Data mining in Learning and using linguistics in classification. This Manuscript covers approaches to General

Human face expression is one of the cognitive activity or attribute to deliver the opinions to others. This paper mainly delivers the performance of appearance based holistic approach subspace methods based on Principal Component Analysis (PCA). In this work texture features are extracted from face images using Gabor filter. It was observed that extracted texture feature vector space has higher dimensional and has more number of redundant contents. Hence training, testing and classification time becomes more. The expression recognition accuracy rate is also reduced. To overcome this problem Symmetrical Weighted 2DPCA (SW2DPCA) subspace method is introduced. Extracted feature vector space is projected in to subspace by using SW2DPCA method. By implementing weighted principles on odd and even symmetrical decomposition space of training samples sets proposed method have been formed. Conventional PCA and 2DPCA method yields less recognition rate due to larger variations in expressions a...

Deriving an effective facial feature from original face images is a vital step for a successful automatic facial expression recognition. In this paper, we proposed a new feature descriptor, Local Principal Texture Pattern (LPTP), for expression recognition. We compute the LPTP feature, at each pixel, by extracting the principal directions of the local neighborhood, and we code the intensity differences on these directions. The mixture of direction and contrast information makes our descriptor robust against rotation and illumination changes.  Consequently, we represent each expression as a distribution of LPTP codes. Our experiments demonstrate the superiority of the proposed feature, on two facial expression databases, over the existing methods.

Traditional art forms are rich sources for understanding and acknowledging cultural elements of a pre-existent society. For a modern society that is engrossed in everything digital, to learn to appreciate such traditions, a transformational effort that alters the availability of these art forms becomes seemingly important. Thus the main objective of this research is to create a virtual environment, in which the common man is adorned as a Kabuki actor, a popular artist from the Kabuki Theatre of Japanese origin, in an attempt to revive fading traditions through the extended reach of computing. We developed KabukiMono+, which tracks user’s face in real-time video stream, and renders proper Kumadori (Kabuki makeup) according to the analysis of facial expression and emotion. Experiments show the robust performance of KabukiMono+.

Human face expression is one of the cognitive activity or attribute to deliver the opinions to others. This paper mainly delivers the performance of appearance based holistic approach subspace methods based on Principal Component Analysis (PCA). In this work texture features are extracted from face images using Gabor filter. It was observed that extracted texture feature vector space has higher dimensional and has more number of redundant contents. Hence training, testing and classification time becomes more. The expression recognition accuracy rate is also reduced. To overcome this problem Symmetrical Weighted 2DPCA (SW2DPCA) subspace method is introduced. Extracted feature vector space is projected in to subspace by using SW2DPCA method. By implementing weighted principles on odd and even symmetrical decomposition space of training samples sets proposed method have been formed. Conventional PCA and 2DPCA method yields less recognition rate due to larger variations in expressions and light due to more number of feature space redundant variants. Proposed SW2DPCA method optimizes this problem by reducing redundant contents and discarding unequal variants. In this work a well known JAFFE databases is used for experiments and tested with proposed SW2DPCA algorithm. From the experimental results it was found that facial recognition accuracy rate of GF+SW2DPCA based feature fusion subspace method has been increased to 95.24% compared to 2DPCA method.

Human face expression is one of the cognitive activity or attribute to deliver the opinions to others. This paper mainly delivers the performance of appearance based holistic approach subspace methods based on Principal Component Analysis (PCA). In this work texture features are extracted from face images using Gabor filter. It was observed that extracted texture feature vector space has higher dimensional and has
more number of redundant contents. Hence training, testing and classification time becomes more. The expression recognition accuracy rate is also reduced. To overcome this problem Symmetrical Weighted 2DPCA (SW2DPCA) subspace method is introduced. Extracted feature vector space is projected in to subspace by using SW2DPCA method. By implementing weighted principles on odd and even symmetrical decomposition space of training samples sets proposed method have been formed. Conventional PCA and
2DPCA method yields less recognition rate due to larger variations in expressions and light due to more number of feature space redundant variants. Proposed SW2DPCA method optimizes this problem by reducing redundant contents and discarding unequal variants. In this work a well known JAFFE databases is used for experiments and tested with proposed SW2DPCA algorithm. From the experimental results it
was found that facial recognition accuracy rate of  F+SW2DPCA based feature fusion subspace method has been increased to 95.24% compared to 2DPCA method.