Local Directional Number Pattern for Face Analysis: Face and Expression

Transcription

1 Local Directional Number Pattern for Face Analysis: Face and Expression Ms. Syed Nahid Kausar & Mr. Suresh S. Gawande 1 Student, Digital Communication, Bhabha Institute of Engg& Tech., nahid_kausar2002@yahoo.com, HOD, Digital Communication Bhabha Institute of Engg& Tech, suresh.gawande@rediffmail.com ABSTRACT LDN encodes the directional information of the face s textures producing a more compact discriminative code than current present methods. With the help of a compass mask that extracts directional information, and compute the structure of each micro-pattern and encode such information using the prominent direction indices (directional numbers) and sign this allows us to distinguish among similar structural patterns that have different intensity transitions. Here we divide the face into several regions to extract the distribution of the LDN features from them. We use it as a face descriptor by concatenate these features into a feature vector. We have performed several experiments in which our descriptor performs consistently under noise, illumination, expression, and time lapse variations. Furthermore, we test our descriptor with different masks to analyze its performance in different face analysis tasks. Keywords: Local directional number pattern, image descriptor, face descriptor, feature, face recognition, expression recognition INTRODUCTION It is a very difficult task to exactly model and animate the facial expressions. In spite of immense efforts in computer hardware and software development, including the improvement of sophisticated algorithms for computer graphics, today still no computational system exists that approximates the performance of humans. The basic purpose of this research is to examine and recognize the essential components relevant to the functions of current methods or techniques in order to create face identification programs. This paper provides a base-lining research for the research of different techniques that advocate its success for each of the method. Local Directional Number (LDN) as a face descriptor for robust face recognition that encodes the structural information and the intensity variations of the face s texture is proposed in this paper. To code the different patterns from the face s textures, directional information is used which is more stable against noise than intensity. 1.1 Proposed Methodology Local Directional Number Pattern (LDN) is a six bit binary code which is assigned to each pixel of an input image that represents the structure of the texture and its intensity transitions. As previous research showed, edge magnitudes are largely insensitive to lighting changes. Hence we create our pattern by computing the edge response of the neighbourhood using a compass mask, and by taking the top directional numbers, that is, the most positive and negative directions of those edge responses. We illustrate this scheme in fig. 1. The estimable information of the structure of the neighbourhood is stipulated by the positive and negative responses, as they reveal the gradient direction of bright and dark areas in the neighbourhood. Thereby, this distinction, between dark and bright responses, grant LDN to differentiate between blocks with the positive and the negative direction swapped (which is equivalent to swap the bright and the dark areas of the neighbourhood by generating a different code for each instance, as shown in the middle of fig. 1) while other methods may mistake the swapped regions as one. Likewise, these transitions occur often in the face, for example, the top and bottom edges of the eyebrows and mouth have different intensity transitions. Hence, it is important to differentiate among them and LDN can accomplish this task by assigning a specific code to each of them.

2 2. LOCAL DIRECTIONAL PATTERN NUMBER SYSTEM Fig-1:Architecture Emotion Expression Recognition The facial expression recognition is performed by using a Support Vector Machine (SVM) to evaluate the performance of the proposed method. SVM is a supervised machine learning technique that implicitly maps the data into a higher dimensional feature space. Accordingly, it finds a linear hyper plane with a maximal margin, to separate the data in different classes in this higher dimensional space. Back Propagation neural networks based algorithm (BPN) BPN is mostly used as a method for recognition process. BPN will be implemented once a face has been detected to identify and recognize who the person is by calculating the weight of the facial information. Basically, BPN imitates human brains biological neuron system. A neuron receives a signal from the previous layer and transmits the signal to all neurons on the next layer. The signal has been multiplied by a separate multi weight value before transmitting the signal to the next layer and the weighted input is summed. BPN is divided into several types such as feed forward neural network, back propagation neural network and Radial Basis Function (RBF) network. These three networks are considered as the most commonly used in BPN. SYSTEM ARCHITECTURE: Index Image Dataset Query Image Extract Facial Features Histogram Generation Expression recognition Retrieve Similar Images

3 2. Modules 1. Local Direction Number Pattern (LDN) 2. Histogram generation for Facial Expression 3. Expression Recognisation 4. Face Retrieval Modules Description 1. Local Direction Number Pattern In this module, LDN is a six bit binary code assigned to each pixel of an input image that represents the structure of the texture and its intensity transitions. Hence we create our pattern by computing the edge response of the neighbourhood using a compass mask, and by taking the top directional numbers, that is, the most positive and negative directions of those edge responses. 2. Histogram generation for Facial Expression In this Module, each face is represented by a LDN histogram (LH). The LH contains fine to coarse information of an image in the form of edges, corners, spots, and other local texture features. To aggregate the location information to the descriptor the histogram encodes the occurrence of certain micro-patterns only, without providing the actual location information. The histogram is generated based on the query image selected from the image dataset. The horizontal axis of the graph represents the tonal variations, while the vertical axis represents the number of pixels in that particular tone. The left side of the horizontal axis represents the black and dark areas, the middle represents medium grey and the right hand side represents light and pure white areas. The vertical axis represents the size of the area that is captured in each one of these zones. Thus, the histogram for a very dark image will have the majority of its data points on the left side and center of the graph. Conversely, the histogram for a very bright image with few dark areas and/or shadows will have most of its data points on the right side and center of the graph. 3. Expression Recognition: We perform the facial expression recognition by using a Support Vector Machine (SVM) to evaluate the performance of the proposed method. SVM is a supervised machine learning technique that implicitly maps the data into a higher dimensional feature space. Consequently, it finds a linear hyper plane, with a maximal margin, to separate the data in different classes in this higher dimensional space. After the his togram identified in the previous module, we extract all the feature automatically and the features are stored separately. Based on the extracted features, the expression is recognized. 4. Face Retrieval: In this module, we retrieve the similar images based on the expression recognized on the previous module. The efficiency of the descriptor depends on its representation and the ease of extracting it from the face. Ideally, a good descriptor should have a high variance among classes (between different persons or expressions), but little or no variation within classes (same person or expression in different conditions). These descriptors are used in several areas, such as, facial expression and face recognition. Fig. Example of pre-processed images of JAFFE Database 3. RESULTS The performance of proposed LDN pattern is tested in the face recognition problem along with the SVM to evaluate the performance of the proposed method with images from the JAFFE and FERET database. These images are cropped and normalized to pixels based on the ground truth positions of the two eyes and a mouth. Every image in our E setup is partitioned into 10x10 subblocks. We used fa image set as gallery image and other four sets of probe images, those are fb (expression variation), fc (illumination variation), dupi (age variation) and dupii (age

4 variation). We compared performance of proposed LDN based method with LBP and PCA and the recognition rate is as shown in Table I. Experimental results reflect that LDN texture description is more robust in lighting condition and aging effects. The proposed method cannot merely recognize face but can recognize with change in pose, age and expression. Method fb fc dupi dupii LDN LBP PCA Table I. Showing experimental results 4. CONCLUSION A unique encoding scheme has been introduced using Local Directional Number (LDN) that takes advantage of the structure of the face s textures and encodes it efficiently into a compact code with the help of SVM. LDN takes advantage of the structure of the face s textures and encodes it efficiently into a compact code. It uses directional information that is more stable against noise than intensity, to code the different patterns from theface s textures. LDN, implicitly, uses the sign information of the directional numbers which allows it to distinguish similar texture s structures with different intensity transitions, e.g., from dark to bright and vice versa. This new code performs with higher accuracy under difference expressions and aging conditions, making the system run reliably in uncontrolled environment. 5. ACKNOWLEDGEMENT First of all I thank the almighty for giving us the knowledge and courage to complete the res earch work successfully. Next I express my gratitude to our respected Assistant Professor MR.Suresh.S Gawande for following us to do research work effectively. Finally, I wish to express my gratitude to my family and my friends, for providing unending support in every aspect possible, to ease my path on this journey. REFERENCES [1]. Adin Ramirez Rivera,Student Member, IEEE,Jorge Rojas Castillo,Student Member, IEEE, and OksamChae,Member, IEEE Local Directional Number Pattern for Face Analysis: Face and Expression Recognition - IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 22, NO. 5, MAY 2013.

5 [2]. R. Chellappa, C. Wilson, and S. Sirohey, Human and machine recognitionof faces: A survey, Proceedings of the IEEE, vol. 83, no. 5, pp , May [3]. W. Zhao, R. Chellappa, P. J. Phillips, and A. Rosenfeld, Face recognition:a literature survey, ACM Comput. Surv., vol. 35, no. 4, pp , Dec [4]. S. Z. Li, A. K. Jain, Y. L. Tian, T. Kanade, and J. F. Cohn, Facialexpression analysis, in Handbook of Face Recognition. Springer NewYork, 2005, pp , / [5]. H. Hong, H. Neven, and C. von der Malsburg, Online facial expressionrecognition based on personalized galleries, in Automatic Faceand Gesture Recognition, Proceedings. Third IEEE InternationalConference on, Apr. 1998, pp [6]. I. Kotsia and I. Pitas, Facial expression recognition in image sequencesusing geometric deformation features and support vector machines, IEEE Trans. Image Process., vol. 16, no. 1, pp , Jan [7]. N. G. Bourbakis and P. Kakumanu, Skin-based face detection-extractionand recognition of facial expressions, in Applied Pattern Recognition,2008, pp [8]. N. Bourbakis, A. Esposito, and D. Kavraki, Extracting and associatingmeta-features for understanding people s emotional behaviour:face and speech, Cognitive Computation, vol. 3, pp , 2011, /s [9]. P. Kakumanu and N. Bourbakis, A local-global graph approach forfacial expression recognition, in Tools with Artificial Intelligence, 2006.ICTAI th IEEE International Conference on, Nov. 2006, pp [10]. A. Cheddad, D. Mohamad, and A. A. Manaf, Exploiting voronoidiagram properties in face segmentation and feature extraction, PatternRecognition, vol. 41, no. 12, pp , [11]. X. Xie and K.-M. Lam, Facial expression recognition based on shapeand texture, Pattern Recognition, vol. 42, no. 5, pp , [12]. P. Ekman, Emotions Revealed: Recognizing Faces and Feelings toimprove Communication and Emotional Life. Times Books, 2003.