Face Recognition System

R. Prema Dr. P. Shanmugapriya

Research Scholar and Assistant Professor Associate Professor

Dept. of CSE Dept. of IT

SCSVMV University SCSVMV University

Kanchipuram,Tamil Nadu, India Kanchipuram ,Tamil Nadu, India.

[e-mail]

Abstract- Security and authorization of person is a crucial part of any industry or organization. There are many techniques are available in the industry .Techniques are Password management system ,finger print, signature matching etc., one of them is face recognition system. It has received substantial attention from researchers in various fields of science such as biometrics and image processing. Face Recognition system. Many researches in face recognition have been dealing

Given aging pattern pair as input, the parallel convolutional neural network is ex-ploited to learn a discriminative space for the verification task.

Both shape and texture of a face contain important information about human age and identity. The cranial size of a face increases quickly as a person grows until 19 years old. After that, the facial texture change becomes the dominant factor for human aging [2]. Wrinkles are deepened at the sides of the eyes, and freckles and aging spots occur on the face skin. However, shape and texture correlate with each other deeply on the face,and are also influenced by other factors, such as pose and illu- mination. This phenomenon makes cross-age face verification an even more challenging problem. Based on the above observations, we extract shape and tex-ture from the faces and model them separately. 68 face landmark points are located by the OMRON face alignment algorithm. Faces are aligned according to the centers of two eyes and that of the mouth. The shape information is represented by the nor-malized coordinates of landmark points on the aligned faces.

To extract shape-free texture, we first calculate the meanshape of all the training images from the dataset.

11: end for

12: Obtain the synthesized aging patterns of all the training subsets. Train the parallel CNNs: on the original images.13: Train the CNN do14: for on the constructed faces of the15: Train the i-th CNN i-th age group.

16: end for to obtain17: Jointly fine-tune the CNNs .the CNN

CONCLUSION AND FUTURE WORK

In this work, we have developed a novel framework DAFV for aging face verification with large gaps. Two modules, aging pattern synthesis and aging face verification, are included in this framework. In the aging pattern synthesis module, we have proposed a novel deep aging-aware denoising auto-encoder ( -DAE) to synthesize the faces of four age groups for the input face of an arbitrary age. In the aging face verification module, given a face pair as the input, each pair of