A Neural Network Approach for Age Estimation Based on Facial Features.

Sneha Thakur and Ligendra Verma

Raipur Institute of Technology, Raipur

*Corresponding Author E-mail: sne_thakur@yahoo.co.in

ABSTRACT:

This paper focuses Automatic age estimation based on face recognition. This approach is based on the Principle Component Analysis (PCA). Eigen face approach is used for both age prediction and face recognition. Face database is created by aging groups individually. The age prediction is carried out by projecting a new face image into this face space and then comparing its position in the face space with those of known faces. After that we find the best match in the related face database, the Eigen face representation of an input image is first obtained. Then it is compared with the Eigen face representation of face in the database. The closest one is the match.

KEYWORDS:

INTRODUCTION:

Face recognition includes one of the biometric systems. Some examples of biometric features of humans are: Signature- studies the pattern, speed, acceleration and pressure of the pen when writing ones signature. Fingerprint- studies the pattern of ridges and furrows on the surface of the fingertip. Voice- studies way humans generate sound from vocal tracts, mouth, nasal cavities and lips. Iris- studies the annular region of the eye bounded by the pupil and the sclera. Retina- studies the pattern formed by veins beneath the retinal surface in an eye. Hand Geometry- measures the measurements of the human hand. Ear Geometry- measures the measurements of the human ear. Facial thermo gram- concerns the heat that passes through facial tissue. Among them face is the most natural and well known biometric. Age prediction is concerned with the use of a training set to train a model that can estimate the age of the facial images. Among the first to research age prediction were, Kwon and Vitoria Lobo who proposed a method to classify input face images into one of the following three age groups: babies, young adults and senior adults⁶. Their study was based on geometric ratios and skin wrinkle analysis. Their method was tested on a database of only 47 high resolution face images containing babies, young and middle aged adults. They reported 100% classification accuracy on these data.

Hayashi focused their study on facial wrinkles for the estimation of age and gender⁸. Skin regions were first extracted from the face images, followed histogram equalization to enhance wrinkles. Then, a special Hough transform, DTHT (Digital Template Hough Transform) was used to extract both the shorter and longer wrinkles on the face. Their experiments were not very successful on the age classification task though, achieving only 27% accuracy of age estimation and 83% on gender classification. It is important to note that they did not mention the size or source of their test to generate their accuracy values. Hayashi also noted the difficulty of extracting wrinkles from females’ ages between 20 and 30 due to presence of makeup⁸. Lanitis empirically studied the significance of different facial parts for automatic age estimation⁹. The algorithm is based on statistical face models. Lanitis claims that introduction of the hairline has a negative effect on the results⁹.His study was limited to subject ranging from 0 to 35 years old, and contained 330 images, of which only 80 were used for testing purposes. Evidently, faces with more wrinkles weren’t used, leaving in doubt his ability to estimate the age of subjects older than 35 years. Some researchers have focused on particular age groups only, while others use an extremely wide classification range. Primarily, due to the lack of a good database, a global age prediction function, covering an extensive range of ages has yet to be developed. J. R. Scolar and P. Navarreto³ proposed a face recognition algorithm based on Eigen space. J. Yang and et al.⁴ introduced the a new approach to appearance-based face representation and recognition. Most of the research in this area is very limited by the size and quality of the database used.

In recent years, a new dimension has been added to the problem of face recognition. Age as an attribute related to human faces is being increasingly studied and there has been a growing interest in problems such as face recognition across ages, automatic age estimation from face images, appearance prediction across aging etc. The research initiatives pertaining to this problem have reached a critical stage and it is essential to streamline future research on this topic in order to make a significant impact on the many day-to-day applications that benefit from solving this problem. In this paper, we attempt to provide a thorough analysis on problems related to facial aging and further offer a complete account on the many research initiatives pertaining to this problem.

A. Motivation characterizing the progressive, but subtle variations in facial appearances as humans age has many significant implications. For instance,

· Homeland security: Face-based authentication systems that typically compare age-separated face images, are bound to benefit from facial aging models and from methods that extract age-invariant signatures from faces. Further, in the absence of such systems, such authentication systems face the cumbersome task of periodically updating large face databases with more recent face images.

· Multimedia: With growing needs to regulate the content viewed by minors on the internet, age-specific human computer interaction systems have found greater relevance in recent years. Hence, methods that perform age estimation are very critical to develop such applications. Further, age-based image retrieval and video retrieval systems are certain to benefit from automatic age estimation systems.

· Missing individuals: Applications that can reliably predict one’s appearance across ages have direct relevance in finding missing individuals.

II RELATED WORK:

The research in age-estimation started in 1990s and up to now, many approaches have been proposed. They typically consist of two main steps: image representation and age prediction. For the image representation, the most common models are Anthropometric model⁸, Active Appearance Model (AAM)⁵, aging pattern subspace⁶, aging manifolds¹, and patch-based model⁷. The final step for age estimation is either the multiclass classification problem or the regression problem.

In 1999, Kwon⁸ measured the changes of face in shapes, e.g. six geometric ratios of key features, to classify faces into appropriate age groups. Drawing inspiration from this work, Ramanathan⁹, Dehshibi¹⁰, later used the geometric ratios of facial features and added information of texture, e.g. wrinkles, in their approaches. Although these approaches achieved low Mean Absolute Errors (MAEs), they can only deal with young ages when the shapes of faces vary largely. Moreover, because of the sensitivity to head pose in the steps of computing geometric ratios in 2D face images, only frontal faces can be used. Adopting the Active Appearance Models (AAMs)¹¹ approach, Lanitis et al.¹², Khoa Luu et al.⁵ used AAM features, which combine both shape and texture information in their ageestimation studies. In 2009, using AAM features extracted from image with 161 landmarks, Ricanek et al.¹¹ developed a multiethnic age-estimation system that can deal with the race problem. Recently, based on the arguments that age information is often encoded by local information, such as wrinkles around the eye corners, other approaches are to divide face images into many sub-regions, extract features from these regions, and then combine them together. Yan et al.⁷ proposed to use Spatially Flexible Patch (SFP) and Gaussian Mixture Model (GMM). B. Ni et al.² developed a technique to extend the human aging image dataset by mining the web resource and then used SFP for representing face images. Suo et al.¹¹ designed a multiresolution hierarchical graphical face model for age estimation. LBP features and Gabor features are also exploited in the work of Günay⁵, and Gao¹⁶. Guo et al.¹⁰, in 2009, investigated the biologically inspired feature (BIF) derived from a feedforward model of the primate visual object recognition pathway – HMAX model. The advantages are that small translations.

Figure 1 System Over view

III PROPOSED WORK:

Face region is extracted from a real image. Firstly noise filtering and image adjusting processes are performed for image enhancing. Thirteen age individual groups are included in a face database. Within a given database, all weight vectors of the persons within the same age group are averaged together. This creates “a face class”. When a new image comes vector is created by projecting it onto the face space. The face is then matched to each face class that gives the minimum Euclidean distance. A ‘hit’ is occurred if the image nearly matches with its own face class and  then the age group that gives the minimum  Euclidean distance will be assumed as the age of the input image. The record of the corresponding person is obtained by comparing with the estimated age group.

PREPROCESSING:

The first step of preprocessing is the extraction. Face region extraction means that image is extracted from input image by using cropping tool. The input color image is converted to gray image and stored in database for processing. The cropped face region and converted gray image are shown in figure .The input image may be current scanned image or realities input image. And them enhancing state occurs. The proposed system allows the free size and format of color image.

Enhancing state includes the noise filtering, gray scale converting and histogram equalization. Histogram equalization maps the input image’s intensity values so that histogram of the resulting Image will have an approximately uniform distribution^3-6. The histogram of a digital image with gray levels in the range [0, L-1] is a discrete function

P(rk)= nk/n

Where, L is the total number of gray  levels , rk is the kth gray level , nk is the number of pixels in the image with that gray level , n  is the total number of pixels in the image and k= 0,1,2…….L-1,p(rk) gives an estimate of the probability of occurrence of gray level rk

Age Prediction System:

The age prediction procedure is described in this section. Features extraction – deals with extracting features that are basic for differentiating one class of object from another. First the fast and accurate facial features extraction algorithm is developed. The training positions of the specific face region are applied. The extracted features of each face in database can be expressed in column matrix shown in figure.

Figure. 3 Feature Extractions

All find the average face for same age group of face imags. The mean face feature for the M face images of each age group can be described as:

The face space is computed from the Euclidean distance of feature points of two faces. The fundamental matrix A is constructed by the difference face space among the input and each face. Then the matrix Ω can be formed by the average face features of the thirteen age groups.

Calculate the covariance matrix Cov = ΩΩT    and then built Matrix L = ΩΩT    to reduce dimension. Find the eigenvector of Cov. Eigenvector represent the variation in faces. Finally age is determined through the minimize face space.

 NEURAL NETWORK STRUCTURE

Figure 4: Neural Network architecture

IV. RESULT:

The performance of age estimation algorithms is normally tested with two different measures: the mean absolute error (MAE) and the cumulative score (CS). However, since the proposed system is concerned with coarse age estimation, these measures could not be adopted to accurately reflect the results since the final output of coarse estimation system is the age range and not the exact age of the human face. Hence, the percentage of accuracy achieved during the experiments was tabulated, charted and presented.

TABLE 1: RESULTS ON BABY TEST CLASS

TABLE 2: RESULTS ON CHILD TEST CLASS

TABLE 3: RESULTS ON ADULT TEST CLASS

CONCLUSION:

Because none of the previous research work is focused on more tuned age ranges as we have done, the comparative study becomes a little bit difficult. Comparing our results with the work presented in⁶ where the authors first classify the input to baby or others and then try to classify others to three age categories. In the first stage the accuracy is very high (up to 99%), and this is simple to obtain because the classification is only to recognize whether the given input is for a baby or others. In the second stage, the new classifier tries to classify the others in one of three age ranges and in this case the performance does not exceed 78.4%. Thus the overall performance is about 88%, but the error rate for the others is very high and if more ages division are considered as we have done, the error rate would be absolutely more. Moreover, the baby age range is just in the range of 1-2 years. This means the 88% does not really scale the overall system performance. If in our case results, we want to consider the others results neglecting the babies, the system performance would be 84% with more correct age ranges which is higher than 78.4% in⁶. This demonstrates that our proposed system can better recognize the others than what is proposed in⁶.

REFERENCES:

1. Fukai, H., Nishie, Y., Abiko , K., Mitsukura, Y., Fukumi, M. and Tanaka, M. "An Age Estimation System on the AIBO", International Conference On Control, Automation And Systems, pp.2551-2554, 2008.

2. Geng, X., Zhou, Z. and Smith-Miles, K. "Automatic Age Estimation Based on Facial Aging Patterns", IEEE Transaction On Pattern Analysis And Machine Intelligence, Vol. 29, No. 12, pp.2234-2240, December 2007.

3. Geng, X., Smith-Miles, K. and Zhou, Z. "Facial Age Estimation by Nonlinear Aging Pattern Subspace", Proceedings Of The 16^th ACM International Conference on Multimedia , pp. 721-724, 2008.

4. Horng, W., Lee, C. and Chen, C. "Classification of Age Groups Based on Facial Features", Journal of Science and Engineering, Vol. 4, No. 3, pp. 183-192, 2001.

5. Cootes , F., C. J. Taylor, D. H. Copper, and J. Graham, "Active Shape Models – Their training and Application", Computer Vision Graphics And Image Understanding, Vol. 61, No 1, pp. 38–59, 1995.

6. Cootes ,F. , Gareth J. Edwards , Christopher J. Taylor, "Active Appearance Models", Proceedings of the 5th European Conference on Computer Vision-Volume II, p.484-498, June 02-06, 1998.

7. http://www.fgnet.rsunit.com .

8. Luu, K., Ricanek Jr., K. and Y. Suen, C. "Age Estimation using Active Appearance Models and Support Vector Machine Regression", Proceedings Of The IEEE International Conference On Biometrics :Theory Applications And System , pp.314-318,2009.

9. Ricanek Jr., K and Tesafaye, T., “ MORPH: A Longitudinal Image Database of Normal Adult Age-Progression", Proceedings of the 7th International Conference on Automatic Face and Gesture Recognition, p.341-345, April 10-12, 2006.

10. http://personalpages.manchester.ac.uk/staff/timothy. f.cootes/data/xm2vts/xm2vts_markup.html

11. http://www.scholarpedia.org/article/Facial_Age_Estimation

Received on 31.10.2011       Accepted on 14.11.2011     

© EnggResearch.net All Right Reserved

Int. J. Tech. 1(2): July-Dec. 2011; Page 90-95