Aging time applications such as identifying missing children, identifying

Aging Face Recognition SystemUsing Local Pattern Selections  Trupti Hankare Shraddha Kasbe UG Student UG Student Department of Computer Engineering Department of Computer Engineering Bharati Vidyapeeth College of Engineering Bharati Vidyapeeth College of Engineering   Pranali Mhatre Pratiksha Musale UG Student UG Student Department of Computer Engineering Department of Computer Engineering Bharati Vidyapeeth College of Engineering Bharati Vidyapeeth College of Engineering  Prof.Madhuri Ghuge Assistant  ProfessorDepartmentof Computer Engineering BharatiVidyapeeth College of Engineering Abstract Aging is the biological processof growing older in a deleterious sense. Automatic Face Recognition is based onrecognizing the faces of same person across different ages which has many realtime applications such as identifying missing children, identifying criminals.The major cause of aging is facial features of same person is changed due toaging process. Intra-user dissimilarity is used to describe the changes thathappened to the same person’s facial features due to aging.

This paper isproposed to solve the problem by introducing an effective method withhierarchical learning model based on Local Pattern Selection (LPS). LPSalgorithm is used for low-level learning visual structures. It minimizes theintra-user dissimilarity. The first level, low level features are extractedusing LPS descriptor and at the second level the output of first level is usedas an input to the second level to produce refined and accurate results. Keywords:Aging Face, Face Recognition, Intra-User Dissimilarity, Feature Descriptor. ________________________________________________________________________________________________________ I.       INTRODUCTION Automatic face recognition is the biometricidentification by scanning a person’s face and matching it against a library ofknown faces.

Best services for writing your paper according to Trustpilot

Premium Partner
From $18.00 per page
4,8 / 5
Writers Experience
Recommended Service
From $13.90 per page
4,6 / 5
Writers Experience
From $20.00 per page
4,5 / 5
Writers Experience
* All Partners were chosen among 50+ writing services by our Customer Satisfaction Team

Automatic age estimation has great interest because of its manyuseful real-world applications-finding missing children, identifying criminalsbased on identity ‘mug shots’ 5 9. Automatic face recognition is aparticularly attractive biometric approach, since it focuses on the sameidentification that humans use primarily to distinguish one person fromanother: their faces. One of its main goals is the understanding of the complexhuman visual system and the knowledge of how humans represent faces in order todiscriminate different identities with high accuracy. Many methods of facerecognition have been proposed. Basically they can be divided into holistictemplate matching based schemes, geometrical local feature based schemes andhybrid schemes of all these types have been successfully applied to the task offace recognition.

Intra-user dissimilarity makes the task very difficult 1. Automatic face recognition consists of subtask in asequential manner- face detection, face segmentation/face normalization andface recognition/verification. By observing features of face, it is categorizedinto three partitions – babies, younger adults, senior adults. First primaryfeatures are extracted and then secondary features deals with wrinkle analysis thathelps in distinguishing senior adult from younger adult 6. The rest of thispaper is organized as follows. In section-II, a detailed explanation of systemis provided. In section-III, this paper presents an efficient featurerefinement framework. Section-IV introduces the simulation results.

The paperis concluded in the section-V.           All rights reserved by 1         AgingFace Recognition System Using Local Pattern Selections (IJSTE/ Volume 2 / Issue 11 / 2018)                         Figure 1:Example faces for one of the subjects in the MORPH Database  II.SYSTEM OVERVIEW In this paper, a two-levellearning model is used to address the problem of intra-user dissimilarity 18.In this system, effective features are learned from low-level microstructuresbased on the feature descriptor called Local Pattern Selection (LPS). LPSselects low level common information between cross-age faces. Then higher levelvisual information is further refined based on output from first level.

Usually, in all image related projects, LBP feature extraction technique isused. Local Binary Pattern (LBP) is simple and accurate but it maximizes theintra-user dissimilarity where as LPS is used to minimize intra-user dissimilarity16 17. III.FEATURE EXTRACTION LPS is the feature extractionmethod used here.

Feature extraction is the technique of extracting features ofthe image like its mean, variance, standard deviation, entropy etc. Now thesefeatures stand instead of the image. 2 A.    Motivation Facial image provides a pool ofinformation. The information regarding age, gender, identity, etc. Agingestimation is the technique of estimating age from face. Aging occurs todifferent person at different rates. Aging rate depends on the intrinsicfactors and extrinsic factor.

Due to this persons of same age look quitedifferently from one another. Intra user dissimilarity is the technical termused here. Intra user dissimilarity is defined as the changes happening to thesame person during the process of aging. Sometimes due to the stressful workingcondition intra user dissimilarity is very high because of that person at twodifferent age appears to be two different individuals.

To overcome thischallenge the LPS is used. It reduces intra user dissimilarity and increasescommon information thus increases the reliability of the system. Theimplementation of feature extraction needs three steps-Pixel feature formation,encoding tree and code assignment 2. B.       Pixel feature formulation Pixel features are the identity of the pixel.

Pixelfeature is a 8 dimensional vector that is formed by comparing central pixelwith adjacent 8 neighbors taken at a radius r. The radius can take values r =(1, 3, 5, 7). Here r is selected as 3. 18       All rights reserved by 2         AgingFace Recognition System Using Local Pattern Selections (IJSTE/ Volume 2 / Issue 11 / 2018)                     Figure 2: Pixel featureformulation C.    Encoding Tree Encoding tree is a binary treemeans zero or one.

Here zero stands for NO and one stands for YES. The encodingtree consist of leaf nodes and internal nodes. Leaf nodes are identified byintegers and internal nodes are identified by attributes and threshold                   .The pixel passes through theencoding tree, it is either routed to left or right 18. Attributes are theelement of the pixel feature that is to be encoded. If the pixel meets thethreshold, it is routed to the right part otherwise to the left part. All thepixel that falls to the same leaf nodes are provided with same codes.

  Figure 3: Encoding tree withinternal and leaf nodes.   D.    Code Assignment The pixels falls under the sameleaf nodes are provided with same code. For example, all the pixel that comesunder the leaf node 5 are assigned with the decimal code 5. 18 Thus minimizesthe intra user dissimilarity which is stated as the major challenge.     All rights reserved by www.ijste.

org 3         AgingFace Recognition System Using Local Pattern Selections (IJSTE/ Volume 2 / Issue 11 / 2018)  IV. SYSTEM ALGORITHM The basic idea of this algorithm is to grow theencoding tree incrementally until it reaches the expected number of leaf nodesL. At each step, the node that maximizes the increasein utility for expansion is selected. The corresponding pixel features of these imagepairs are denoted as:  A = {(x-nm, y-nm)|m=1,…,M; n=1,…,N} (1)  Where, M = H _W is the total number of pixels in animage.

 We say x-nm and y-nm are matching pixel features as they belong to the same pixel locationin the images of the same subject 18. For any encoding tree T of L leafnodes, suppose the T assigns each pixel feature in A with a code based on whichleaf node that pixel finally reaches, resulting in an encoded set:  C= {(u-nm, u-nm)|m=1,…,M; n=1,…,N} (2) Let’s denote the tree with K leaf nodes (1 _ K

Output: Encoding tree T. /* Pixel features extraction. */ begin Convert images into a set of pixel features asdescribed in Eqn (1): A  = {(x-nm, y-nm)|m=1,…,M; n=1,…,N} /* Encoding tree initialization.

*/ begin Initialize encoding tree T by adding one leaf nodew, whose indices of support pixel features are: S1w = (I1(1),….,In1(1) | Ii(1) €1,…,M*N}S2w = (I1(2),….,In2(2) | Ii(2) €1,…,M*N}w.a?0,w.t?0 and w.

?u?-inf. /* Encoding tree learning */ begin for step = 2? L do for each leaf node w do if w.?u ? -inf then /* Node has been evaluated. */ continue; else for k = 1?8 do for z = min(S1w,S2w) ? max(S1w,S2w)do   All rights reserved by 4         AgingFace Recognition System Using Local Pattern Selections (IJSTE/Volume 2 / Issue 11 / 2018)  Evaluate increase of utility ?u with attribute = kand threshold = z.

 Let maximum ?u* is achieved at (k*; z*). Update: w.?u = ?u*, w.

a ?k*, w.t? z*. Let w*has maximum _u over all leaf nodes. Split w* into two children nodes l and r. l.a ?0,l:t   0, and l.?u ? -inf.

 r.a ?0, r:t 0, and r.?u ? -inf. Update S1l ,S2l, S1r ,S2r based on Eqn (3),(4). Assign distinct codes to leaf nodes, and return T.18  IV. LPSBASED FEATURE EXTRACTION The Algorithm 1 learns an encoding tree that encodes given image byconverting each pixel into decimal codes based on the leaf node that pixelreaches in the tree. In this part, we briefly introduce how we extractover-completed features based on encoded images.

The techniques we use includemultiple scaling and dense sampling 8. Specifically, we first train multipleencoding trees based on different sampling radii (e.g.1; 3; 5; 7) as illustratedin Figure 2. Then for each encoded image, weextract local features by calculating the histograms of small patches formed bydividing image into overlapping (with overlapping factor 0:5) fixed size (e.g.16 _ 16) areas.

The final features of an image are formed by concatenatinglocal features at all sampling radii. V.INFORMATION REFINEMENT Feature extraction technique LPSis used to collect low level visual structures. The features collected usingLPS are of high dimension due to the employment of multiple scaling and densesampling technique. The high dimension features arises problems in storage andmatching. Same features may be there for many times. Redundant data and noisemay mislead to wrong decision.

So it is essential to filter out these redundantdata and noise, thus making set of features more precise. Filtering points tothe need of refinement technique 2. The most popular refinement techniquecommonly used in image processing is Fisher linear discriminant analysis oruniversal subspace analysis (USA). It deals with the eigen decomposition of amatrix of size Min (N, D) where N is the number of training image pairs and D is the dimension of features. The time complexity of eigen decomposition is around min {O(N3), O (D3)}.Thus making it not suitable for very large values of N and D 218.

Here this part presents a refinement technique that successfullyeliminates redundant data and noise, thus provides crucial data for theclassification. This refinement technique has two stages. First one is thebootstrap aggregating and the second is random subspace classifiers. A.   Bootstrap aggregating: Divide Data by Sampling Bootstrap aggregating isotherwise known as bagging.

The one line definition of the technique isdividing data by sampling. The data set include pictures of 10,000 persons thatmeans N = 10,000, N is defined as the no: of training image pairs. In the dataset two photographs of same person taken at a large age gap are included. .Bagging technique is used to select a series of subset M (M

Let M =2000, N = 10,000 and K = 20, K is the number of the training set. From 10,000person set 20 training sets are formed each training set photographs of 2000person. Same persons are included in many training set to check whether all theclassifiers produces the same result. When the value of N increases, the value ofK also increases. The increase in K leads to an increase in computational time. B.       RandomSubspace: Divide Features by Sampling Bagging helps in limiting the computation cost. Forfurther improvement in stability random subspace classifiers are used.

 Random subspace algorithm is used to make a randomselection of samples from the training subsets (e.g. take 5K out of 100K). Universal Subspace Analysis (USA) is used as baseclassifiers. 20    All rights reserved by www.ijste.

org 5         AgingFace Recognition System Using Local Pattern Selections (IJSTE/ Volume 2 / Issue 11 / 2018) C.   Discussion The second level classificationframework is designed to be scalable. As described in Section III-A, as thenumber of training samples increases, the training time increases linearly.Specifically, for the Morph 2 dataset, the major computational cost involves2000x2000 square matrix eigen decomposition.

There are K=20 of thesedecompositions, and these decompositions can be computed in parallel 18. On a6-core machine it takes 3.64 seconds for each of the decomposition, and 74.31seconds in total for the second level (including other overhead).

This paper isfocused on training time analysis, since for testing, it only involvesmatrix-vector multiplication, and thus it is very fast. VI.CONCLUSION Automatic age estimation alongwith personal information gathering can be used as a helping hand to determinethe criminals more efficiently and accurately by comparing the data set ofcriminals with mug shot.

Aging estimation is the process of determining thespecific age or age range of a subject based on a facial image. Automatic ageestimation has growing interest due to its applications in real life. New andnew projects are developing on the topic automatic age estimation, but still itis not that easy to determine age from face. Because aging rate for differentperson will be different. Aging rate greatly depend on the intrinsic factorsand extrinsic factors.

Intrinsic factors are genetic factors and extrinsicfactors are lifestyle, expression, and environment. As a result, differentpeople with the same age possess quite different appearances due to differentrates of facial aging. To overcome this problem here LPS feature extractiontechnique is used. LPS is taken because of its ability to minimize intra userdissimilarity and at the same time maximizes common information, thusincreasing the accuracy of the system 2.