IEEE International Conference on Computer, Communication, and Signal Processing (ICCCSP-2017) 978-1-5090-3716-2/17/$31.00 ©2017 IEEE Facial Recognition using Histogram of Gradients and Support Vector Machines J. Kulandai Josephine Julina Department of Information Technology SSN College of Engineering Chennai, India josephinejulinajk@ssn.edu.in T. Sree Sharmila Department of Information Technology SSN College of Engineering Chennai, India sreesharmilat@ssn.edu.in Abstract— Face recognition is widely used in computer vision and in many other biometric applications where security is a major concern. The most common problem in recognizing a face arises due to pose variations, different illumination conditions and so on. The main focus of this paper is to recognize whether a given face input corresponds to a registered person in the database. Face recognition is done using Histogram of Oriented Gradients (HOG) technique in AT & T database with an inclusion of a real time subject to evaluate the performance of the algorithm. The feature vectors generated by HOG descriptor are used to train Support Vector Machines (SVM) and results are verified against a given test input. The proposed method checks whether a test image in different pose and lighting conditions is matched correctly with trained images of the facial database. The results of the proposed approach show minimal false positives and improved detection accuracy. Keywords- AT & T facial database; Face recognition; Feature extraction; Histogram of Oriented Gradients; Support Vector Machine I. INTRODUCTION The face is an important identity of a person. It is obvious that humans are experts in recognizing people at a farther distance, predicting the identity of a person in more prominent occlusions and in different lighting conditions. However, for the system, it requires many iterations and tuning of threshold parameters to recognize a person and to reject appropriately when a wrong input is given. The location of facial components namely eyes, nose, and mouth constitute prime landmarks in marking the presence of a face in an image. However, this process would be made difficult when a person tends to exhibit different expressions and pose variations. Some of the commonly encountered challenges in facial detection and recognition include variations in lighting conditions, occlusion, wearing of spectacles, having more facial hair and so on. Certain techniques like template matching methods [19], geometry based approaches [1], appearance-based models [3] are adopted to overcome such challenges. Face detection is the first and foremost step in identifying the presence of faces in an image. Then face recognition or verification is carried out which checks whether a given test input containing face is matched with already available faces stored in the database or face gallery. The details about the face and its different facial components can be obtained based on features which are chosen in such a way that they are robust to noise and different orientation. Most of the facial recognition problems deal with feature extraction and machine learning techniques. Facial recognition tasks are performed in many areas of image and vision applications where security is more focused without any second thought of compromising it. Many applications deal with mug shot images which are used to recognize faces of any matched criminals in the database. The remaining sections of the paper is organized as follows: Section II provides details about related work carried in recognition of faces and feature extraction. Section III provides details about the architecture of the proposed method for detecting whether an image corresponds to matched face in the gallery. Section IV provides qualitative results obtained after testing with images of AT & T facial database [16]. The system is able to detect 90% of faces over 41 test images resulting in few false positives. Section V provides conclusion and future work followed by references. II. RELATED WORK The well-known datasets available to detect and recognize faces and facial expressions [18] includes CVL [21], CMU pose, illumination & expression (CMU - PIE) [25], FERET [8], extended M2VTS dataset [17], Cohn-Kanade (CK) [24], JAFFE dataset [20] and so on. This paper uses AT & T facial database [16] and real time images to recognize faces based on feature extraction using HOG [10]. Faces can be detected most generally using Viola-Jones algorithm [9]. Facial features being extracted should be invariant to pose variations. Such invariant features can be obtained based on Active Shape Model (ASM) [4] and Active Appearance Model (AAM) [12]. Features of an image include edge details, texture, color information, corners, and interest points of regions that constitute what is known as the Region of Interest (ROI). The feature space can be reduced by identifying dominant and unique feature using Principal Component Analysis (PCA) [5] technique or by means of distance metric like Euclidean [2]. Some of the well known facial recognition algorithms use Gabor filters [30], LBP (Local Binary Pattern) [30] and Linear Discriminant Analysis (LDA) [30]. The raw pixel information does not discriminate Authorized licensed use limited to: Sri Sivasubramanya Nadar College of Engineering. Downloaded on January 04,2022 at 09:50:53 UTC from IEEE Xplore. Restrictions apply.