Face Recognition: Identifying Facial Expressions Using Back Propagation Manisha Agrawal 1, Tarun Goyal 2 and Harvendra Kumar 3 1 B.Tech CSE Final Year Student, SLSET, Kichha, Distt: U. S, Nagar, Uttarakhand, INDIA 2 Assistant Professor, CSE Department, SLSET, Kichha, Distt: U. S. Nagar, Utarakhand, INDIA 3 Assistant Professor & Head, CSE Department, SLSET, Kichha, Distt: U. S. Nagar, Utarakhand, INDIA ABSTRACT Face recognition is a biometric which uses computer software to determine the identity of the individual. In order to recognize a person, one commonly looks at faces, which differentiate one person to another. Face recognition records the spatial geometry of unique features of the face. Face recognition has always been a very challenging task for the researches. Facial recognition is a form of computer vision that uses faces to attempt to identify a person or verify a person s claimed identity. This paper presents basic algorithm for face recognition and identifying facial expressions using back-propagation of neural network. Keywords: Face Recognition, Eigenface, Principal Component Analysis (PCA), Back-Propagation 1. INTRODUCTION Face Recognition is used to search for other images with matching features [5]. Face recognition falls into the category of biometrics which is the automatic recognition of a person using distinguishing traits [8]. There is a five step process involved with the system developed by Turk and Pentland. First, the system needs to be initialized by feeding it a set of training images of faces. This is used these to define the face space which is set of images that are face like. Next, when a face is encountered it calculates an eigenface for it. By comparing it with known faces and using some statistical analysis it can be determined whether the image presented is a face at all. Then, if an image is determined to be a face the system will determine whether it knows the identity of it or not. The optional final step is that if an unknown face is seen repeatedly, the system can learn to recognize it. Facial recognition is including five steps to complete their process. Step1: Acquiring the image of an individual s face Two ways to acquire image: 1) Digitally scan an existing photograph 2) Acquire a live picture of a subject. Step2: Locate image of face Software is used to locate the faces in the image that has been obtained. Step3: Analysis of facial image Software measures face according to is peaks and valleys; focuses on the inner area of the face identified as the golden triangle, valleys are used to create a face print with their nodal points. Step4: Comparison Face print created by the software is compared to all face prints the system has stored in its database. Step5: Match or no match Software decides whether or not any comparisons from step 4 are close enough to declare a possible match. Facial recognition utilizes distinctive features of the face - including the upper outlines of the eye sockets, the areas surrounding the cheekbones, the sides of the mouth, and the location of the nose and eyes - to perform verification and identification. 1.1 Functions of Face Recognition Face detection: Face detection and indication of any facial zones that are opposite in various guidelines in complex scene. Facial pose estimation: Estimation of the angle to which a face is twisted. Facial part detection: The identification of the positions of face parts for example the centre of eyes, tip of nose, and corners of the jaws. Volume 1 Issue 6 December 2013 Page 7
Facial trait classification: The classification of faces by color, gender, civilization, age, appearance and other character. Face identification: The identification of persons by comparisons with registered people. The following paper discuss about face recognition methods in section II, section III discuss about facial expression recognition ; Section IV explains the training of back propagation neural network that has been described in Section II & Section V is a conclusion phase. 2. EIGENFACE-BASED RECOGNITION The first task of the system is to locate the face (or faces) within the image. Then the facial characteristics are extracted. Facial recognition technique is developed using two metrics: facial metrics and eigenfaces. The face of a human has several features such as, mouth, eyes, nose, eyebrows, and forehead. Each of this features has a unique shape and a unique pattern, hence, many experiments have been reported in extracting facial feature for recognizing facial expression. We extract the features toward forehead wrinkle, mid forehead wrinkle, cheek wrinkle, and mouth length as seen in Figure 1. Figure 1 Parts of Facial Feature Extractation Facial metrics technology relies on the measurement of the specific facial features (the systems usually look for the positioning of the eyes, nose and mouth and the distances between these features which is usually called as golden triangle as shown in Figure 2. Figure 2 Identification of Golden Triangle Eigenfaces is the name given to a set of eigen vectors when they are used in the computer vision problem of human face recognition. Eigenfaces are calculated from the training set and stored. An individual face can be represented exactly in terms of a linear combination of eigenfaces [2]. The face can also be approximated using only the best M eigenfaces, which have the largest eigen values. It accounts for the most variance within the set of face images. Best M eigenfaces span an M-dimensional subspace which is called the "face space" of all possible images. For calculating the eigenface PCA algorithm was used [3], [4]. It includes the calculation of the average face (φ) in the face space and then further computes each face difference from the average. The difference is used to compute a Volume 1 Issue 6 December 2013 Page 8
covariance matrix (C) for the dataset. The covariance between two sets of data reveals how much the sets correlate. Based on the statistical technique known as PCA, the number of eigenvector for covariance matrix can be reduced from N (the no. of pixels in image) to the number of images in the training dataset. Eigen vectors so formed are then projected into Eigen faces as shown in Figure 3. Figure 3 Extraction of EigenFaces using PCA Eigenfaces FR method is based on categorizing faces according to the degree of fit with a fixed set of 150 master eigenfaces. This technique is similar to police methods that are used in creating a portrait; the only difference is that image processing is automatic and based on a real picture. Every face is assigned a degree of fit to each of the 150 master eigenfaces, only the 40 template eigenfaces with the highest degree of fit are necessary to reconstruct the face with the accuracy of 99%. Better results can be achieved if the operator is able to tell the system exactly where the eyes are positioned. The eigenface technique is simple, efficient, and yields generally good results in controlled circumstances [5]. The system was even tested to track faces on film. There are also some limitations of eigenfaces. There is limited robustness to changes in lighting, angle, and distance [8]. 2D recognition systems do not capture the actual size of the face, which is a fundamental problem [6]. 3. FACIAL EXPRESSION RECOGNITION Back-propagation algorithm to recognize of facial expression with feed-forward architecture. Back-propagation is a systematic method of training multilayer artificial neural networks [9]. It can be used to model complex relationships between inputs and outputs or to find patterns in data. The back-propagation of feed-forward architecture [10] is designed based on facial features extracted as illustrated in Figure 4. Figure 4 Architecture of feed-forward back-propagation neural network for facial expression recognition It consists of: An input layer containing four neurons representing input variable to the problem that is extracted data from the forehead wrinkle, the mid forehead wrinkle, the cheek wrinkle, and the mouth length; One hidden layers containing one or more neurons to help capture the nonlinearity in the data; and An output layer containing six nodes representing output variable to the problem that is facial expressions: anger, disgust, surprise, happiness, sadness and fear. The neurons between layers are fully interconnected with weight vij and wij. 4. TRAINING OF THE ARCHITECTURE BY BACK-PROPAGATION NEURAL NETWORK The training of a network by back-propagation neural network involves three stages: Volume 1 Issue 6 December 2013 Page 9
Stage 1: Feed-forward of the input training pattern Stage 2: Calculation and back-propagation of associated error Stage 3: Adjustment of the weights Data is fed forward from the input layer, through hidden layer, to output layer without feedback. Then, based on the feed forward error back-propagation learning algorithm, back propagation will search the error surface using gradient descent for point. Thus, error is computed, and then the weights for all layers are adjusted simultaneously to minimize the error. In many neural network applications, the data (input or target patterns) have the same range of values. We use the binary sigmoid function, which has range of (0, 1) and is defined as f(x) = 1/ (1 + exp (-x)), that's why the data is also represented in binary form or has range of 0-1. Table 1 shows the data of training pairs (input and target patterns) in back propagation of neural network. We use two pairs of training input data for each of six output expressions. The first row is for neural expression. Table 1: Data of Training Pairs in Back-PropagationNetwork 5. Conclusion Face Recognition is a rapidly evolving technology that is being widely used in forensics, security; prevent unauthorized access in bank or ATMs, in cellular phones, smart cards, PCs, in workplaces, and computer networks. This Paper explained various metrics needed for face recognition approach, that is facial metrics and eigen faces using PCA. A simple method for facial expression recognition using back-propagation is also presented. Through the experimental results, it has been found that the expression of sadness and disgust are more difficult than the others to recognize. Face recognition has become an important issue in many applications such as security systems, credit card verification, criminal identification etc. Even the ability to merely detect faces, as opposed to recognizing them, can be important. References [1] Renu Bhatia Biometrics and Face Recognition, Volume 3, Issue 5, May 2013, International Journal of Advanced Research in Computer Science and Software Engineering [2] Mayank Agarwal, Manish Kumar, Nikunj Jain, himanshu Agarwal Face Recognition using Principle Component Analysis, Eigenface and Neural Network [3] Kirby, M., and Sirovich, L., "Application of the Karhunen-Loeve procedure for the characterization of human faces", IEEE PAMI, Vol. 12, pp. 103-108, (1990). [4] S. Gong, S. J. McKeANNa, and A. Psarron, Dynamic Vision, Imperial College Press, London, 2000. [5] Ryan Johnson, Kevin Bonsor, "How Facial Recognition Systems Work," How Stuff Works, 2007. [6] Trina D. Russ, Mark W. Koch, Charles Q. Little, "3D Facial Recognition: A Quantitative Analysis," 38th Annual 2004 International Carnahan Conference on Security Technology, 2004. [7] Bonsor, K. "How Facial Recognition Systems Work, Retrieved 2008-06-02. [8] John D. Woodward, Jr., Christopher Horn, Julius Gatune, Aryn Thomas, Biometrics, A Look at Facial Recognition, RAND, 2003.M. Young, The Technical Writer's Handbook. Mill Valley, CA: University Science, 1989 Volume 1 Issue 6 December 2013 Page 10
[9] S. Rajasekaran, G.A. Vijaylakshmi Pai Neural Networks, Fuzzy Logic, and Genetic Algorithms-Synthesis and Applications, PHI, Fifteenth Printing, July 2011 [10] Neeraj Shukla, Anuj Kumar, Using Back-Propagation Recognition of Facial Expression JECET; December 2012 -February 2013;Journal of Environmental Science, Computer Science and Engineering & Technology AUTHOR Manisha Agarwal pursuing B.Tech degree in Computer Science Engineering from SLSET, Kichha, Uttarakhand, INDIA. She is doing research in the Field of Neural Networks. She is also engaged in many real time live projects for SLSET, Kichha, Uttarakhand, INDIA. She has developed Student record management system for Surajmal College of Engineering & Management, KIchha, Uttarakhand, INDIA. Tarun Goyal received his M.Tech degree in Computer Science Engineering from KEC, Dwarahat, INDIA in 2012 & B.Tech degree in Information Technology from GECB, Bikaner, INDIA in 2010. He is presently working as Assistant Professor in Computer Science department with SLSET, Kichh, Uttarakhand, INDIA. He has published many research papers in the field of Cloud Computing, & Web Technology. He has also developed cloud based application Cloudtarun. He is in Editorial Board of many Journals. (Total Paper: 21, including IEEE, Elsevier, IJCA, IJACR etc). Harvendra Kumar received his ME degree in Software Engineering from Thapar University, Patiala, Punjab, INDIA in 2009 & B.Tech degree in Information Technology from UP Tech University UP, INDIA in 2004. He is presently working as Assistant Professor in Computer Science & IT department with SLSET, Kichh, Uttarakhand, INDIA. He has published many research papers in various field of research upto yet. He has also guided many students in B.Tech Final Year Project. Volume 1 Issue 6 December 2013 Page 11