Study on Face Recognition Algorithm

Abstract

Biometrics authentication is started to be used worldwide since the security it provides is of greater extends. Face recognition is an important method of such an authentication. Face recognition can be carried out by many algorithms such as PCA, LDA, ICA, BIC, AAM etc. In this paper, a detailed study about these algorithms is performed. The study deals with the definition part, methodology and the classification as well. Biometrics or biometric authentication is a way of identifying humans using their characteristics. It is a method of identification and access control. Biometric identifiers are categorized into physiological characteristics which are related to the shape of the body including fingerprints, face recognition ,iris recognition etc. and behavioral characteristics related to the pattern of behavior of person including typing rhythm, voice etc. Universality, uniqueness, measurability, permanence, performance, acceptability and circumvencce are the seven weighted factors for determining the biometrics for use in a particular application. Universality means that the person should posses the trait. Uniqueness means the trait should be unique. Permanence is the manner in which the trait varies over time. Measurability is the ease of acquisition. Acceptability is how others accept the technology. Performance refers to the accuracy and speed and robustness of the technology. Circumvention is the easiness in imitating the trait. A face recognition system compares the video or image of a person in database to the input captured. Comparison is done on the structure, shape and proportions of face. Distance between the eyes , jaw, mouth and nose. Several pictures of the subject at different angle with different facial expressions are enrolled to the database at the training phase. Identification and verification is performed by the subject standing in front of the camera for a few seconds. This image is compared to the one previously added. Different face recognition techniques are feature extraction, neural network, automatic face procession Eigen faces etc. Biometric techniques can be used for surveillance purpose. I.e. Face recognition can be done from a particular distance without having any contact with the subject so that they are unaware about it. Among several information in human faces, its efficient to choose the unique information while discarding other irrelevant information for making the system efficient. In this paper, a study on different face recognition algorithms is performed. These algorithms include PCA, LDA, BIC, ICA, AAM and Gabor Vectors.