Age Estimation

In recent trends estimating human age automatically from face images have lot of potential in real world applications, such as vending machine, security system/network access control, human computer interaction, multimedia communication, video surveillance, customer profiling, demographic statistics collection etc. Security applications have atmost importance in this area.The biometric features of each human being are unique. Age estimation is determines a person’s age or age group using facial images. A database of facial images is trained to extract features using algorithms such local binary patterns[LBP], active shape models[ASM], histogram of oriented gradients[HOG].Age estimation can be done using 3 age groups: child, adult, senior. Age

[5] investigated the biologically inspired features (BIF) for human age estimation from faces. Manifold learning techniques are adopted to embed face images into a low-dimensional aging manifold. The age manifold based regression [5] produces a MAE of 5.07 years on the FG-NET aging database. More recently, Cao et al. [10] argued that these local descriptors use manually designed encodings, and it is difficult to get an optimal encoding method. As shown in [10], the existing handcrafted codes are unevenly distributed, and some codes may rarely appear in face images. This means that the resulting code histogram is less informative and less compact. They used a learning-based encoding method, which adopts unsupervised learning methods to encode the local micro-structures of the face into a set of discrete codes. With Principal Component Analysis (PCA) and normalization, their learning-based descriptor achieves superior performance on face verification.

II. LITERATURE REVIEW

The topic of face image processing has been active and much interest has been shown. Face image processing is a broad topic and has been active for many years. There have been various contributions and different approaches that attempt that to solve or improve age estimation.

Ranjan Jana et al. [1] provided a methodology to estimate the real age of a human by analyzing wrinkle area of face images Wrinkle geography areas are detected and wrinkle features are extracted from face image.

In classification different classifiers are trained and tested by different extracted features. Different classifiers are combined to minimize the classification error rate. In pattern recognition the k-nearest neighbor algorithm (K-NN) is more widely used method for classifying objects based on majority votes of its neighbors.There are various Classification methods which are Support vector machines/regressors, Neural networks. PLS and CCA subspace learning algorithms also used for