Abstract – Humans can be individualized based on their facial characteristics. However,identifying a face becomes a challenging taskdue to the complexities involved. The Visual models of face include many parameters. In this paper different algorithms and parameters are studied and computed for real time results. The combined PCA and geometric feature based algorithm is used to recognize age, gender and facial information. These holistic and geometric methods can be combined for decision making to reduce the error in recognition system.
Keywords –Face Recognition, PCA, Age and Gender classification, Eigen faces, Eigenvalues, Shape transformation, Decision method. I. INTRODUCTION
The face among other biological features is the most emerging research field in biometrics. The growing demands for surveillance, security and identification in societies have raised there search in this field. The change in structure and features on face makes it challenging for programmers and developers.
MD Malkauthekar and SDSapkal  showedexperimental analysis of classification of facial images. G Mallikarjuna Rao  proposed a Neural Network-basedfrontal face detection system to classify the gender based on facial information. Anil Kumar Sao and B. Yegnannarayna  discussed analytic phase based face recognition to resolve the issue of illumination variation using trigonometric functions. Whereas in this paper the information theory is used on face images to decompose it into smaller set of Eigenfaces and principal component analysis is used to create an initial training set. The input image is then compared with the database of Eigenface sets by projecting it on the subspace of the Eigenfaces. The Euclidean distance is calculated and is compared with set threshold for recognition. The PCA gives efficient and good results in lower dimensional space . The geometric features on faces like distance between eyebrow, eyes and size of lips and nose etc. are located and computed using canny edge operator. The gender and age is determined considering region, size and density of mustache, wrinkle and total number of pixels of skin color.
The approaches in the paper for biometric face recognition enhances the existing processes by providing more accuracy and the new added decision method gives strength to used algorithm in identifying the person from the databases. The problems of the variation of light, size, partial occultation, partial noises and illumination during image capture are reduced to an extent during preprocessing of image for recognition system.
Face recognition system is a subset method of pattern recognition. The input face image is compared with the existing database and output is either a known or unknown identification. The outline for face recognition system is given in Figure 1. There are six main functional blocks.
The acquisition module
The face image as the input to face the recognition system is given in this module in standard image formats like jpeg and bitmap etc.
The preprocessing module
The preprocessing module enhances the quality of image by performing certain operations on it. The pre-processing steps implemented are as follows:
(i) Color Conversion: The red, green and blue components of a color (RGB) image at a specific spatial location are MxNx3 array. Three dimensional RGB is converted into two dimensional gray scale images for easy processing of the face image.