Feature Extraction in Face Recognition: A Review

Feature Extraction in Face Recognition: A Review

Gurpreet Kaur, Monica Goyal, Navdeep Kanwal

Abstract: Face recognition is a type of biometric software application by using which, we can analyzing, identifying or verifying digital image of the person by using the feature of the face of the person that are unique characteristics of each person. These characteristics may be physical or behavior. The physiological characteristics as like finger print, iris scan, or face etc and behavior characteristics as like hand-writing, voice, key stroke etc. Face recognition is very useful in many areas such as military, airports, universities, ATM, and banks etc, used for the security purposes. There are many techniques or algorithms that are used for feature extraction in face recognition. In this paper, we make a review of some of those methods which are used for the face recognition that are Principal Component Analysis (PCA), Back Propagation Neural Networks (BPNN), Genetic Algorithm, and LDA, SVM, Independent Component Analysis (ICA)

Get Help With Your Essay

If you need assistance with writing your essay, our professional essay writing service is here to help!
Find out more about our Essay Writing Service

Each method has different -2 functions that are used for the face recognition. Dimensionality is reduced by using theEigen face approach or PCA, LDA to extract the features from images. Genetic Algorithm is based on feature selection and Back propagation Neural Network (BPNN) is used for the classification of face images. Keywords: – Face recognition, PCA, LDA, Features extraction, BPNN. Through this paper my aim to explain all algorithm and compare, that all algorithms that are used for feature extraction in face recognition.

1. INTRODUCTION

Face recognition process is used to recognize a person by using some features of that person and match that feature with digital image. The features that are extracted for the face recognition are eyes, nose, skin, iris, fingeretc. Some authors define it as “Face recognition is considered to be an important part of the biometrics technique or software application by using which, we can analyzing, identifying or verifying digital image of the person by using the feature of the face of the person that are unique characteristics (physical or behavioral characteristics) of each person”[1].

Applications of face recognition: – There are many applications of face recognition such as

• It used for person verification.
• It is used to check criminal records.
• The main purpose of it is security and and for security it has many applications such as Air ports, Military bases, Government offices, where we use it.
• The financial services industry used it to revolves around the concept of security
• It is used to identification documents, (Passports, Driver’s licenses, and ID Cards) of any person, in face recognition.
• Other many applications of it.

Face recognition method consists of three components: –

1. Face detection: – Face detection is used to detect the location of any faces within an image.
2. Feature extraction: – Feature extraction is consisting of segmentation, image rendering and scaling of face are prepared for identification.
3. Face identification:- Mathematical techniques are used for face identification on the features in the facial area of an image.

2. STEPS INVOLVED IN FACE ECOGNITION PROCESS TO EXTRACT FEATURES:-

1. The first step in the face recognition is to select the digital image on which, the algorithms are applied .The selected image is known as the input image which is shown in the given diagram (Fig1).
2. The second step is Face detection is concerned with finding whether or not there are any faces in a given image and if the face is present, return the image location and content of each face.
3. After the face detection, the face extraction process is done, to provide effective information that is useful for distinguishing between faces of different persons.
4. The face recognition is the last process that used the many algorithms to recognize the face in the image.
5. And in last the result is given that is an image is identify or verified.

The algorithms that are used for face recognition are Principal Component Analysis (PCA), Linear discriminant analysis (LDA), Back Propagation Neural Networks (BPNN), Genetic Algorithm, and Support Vector Machine (SVM). Linear discriminant analysis (LDA) and Principal Component Analysis (PCA) both are powerful algorithms used to reduce the dimensionality and extract the features from the image in face recognition technique. The major difference between these two methods is that LDA algorithm selects features that are most effective for class separability whereas PCA selects features important for class representation[].

Figure 1: A Systematic diagram of generic face recognition.

And in now days, a three dimensional facial recognition is the new method in the market and 3-D sensors are used to capture the image for three dimensional facial recognition.

3. METHODS FOR FEATURE EXTRACTION

The number of methods that are used for the Face Recognition such as Principal Component Analysis (PCA), Linear discriminant analysis (LDA), Back Propagation Neural Networks (BPNN), Genetic Algorithm, and Support Vector Machine (SVM). Each method has it different functions and formulas that are used on the database images for experiments. Some of methods from these are given below with their steps.

2.1 Principal Component Analysis (PCA)

The main purposes of a principal component analysis are the analysis of data to identify patterns and finding patterns to reduce the computational complexity of the dataset with minimal loss of information. They are able to provide higher accuracy in extracting facial features for human face identification. PCA finds a linear projection of high dimensional data into alower dimensional subspace. The Steps for the PCA algorithm are the following:-

1. Take the whole dataset consisting ofd-dimensional samples ignoring the class labels
2. Calculate thed-dimensional mean vector.
3. Compute the covariance matrix of the whole data set.
4. Calculate the eigenvectors of the covariance matrix and corresponding eigenvalues
5. Sort the eigenvectors by decreasing eigenvalues and choosekeigenvectors with the largest eigenvalues to form adxkdimensional matrixW.
6. Use thisdxkeigenvector matrix to transform the samples onto the new subspace.This can be summarized by the mathematical equation:

2.2 Discrete Cosine Transform

Discrete Cosine Transform (DCT) isa holistic approach that is used for Local and Global Features involves recognizing the corresponding face image from the database in Face Recognition. DCT is an accurate and robust face recognition system that used certain normalization techniques [11]. The image is cropped from the original image and eliminating the back ground the size of image is 128 × 128 pixels. All images in the database are gray level images. Local features are extracted from the image such as eyes, nose and mouth and DCT is applied on these extracted features. There are some steps that are used in DCT algorithm for encoding operation are following:

1. The size of the input image is N × M.
2. F (i, j) is the intensity of the pixel at x(i, j).
3. F (u, v) is the DCT coefficient for the pixel at x (i, j).
4. For most images, much of the signal energy lies at low frequencies. These appear in the upper left corner of the DCT.
5. Compression is achieved since the lower right values represent higher frequencies, and are often small enough to be neglected with little visible distortion. The output array of DCT coefficients contains integers; these can range from -1024 to 1023.

2.3 Linear Discriminant Analysis (LDA)

Linear Discriminant Analysis (LDA) are twopowerful tools used for data reduction and feature extraction in the appearance-basedapproaches.LDA (also known as Fisher’s Discriminant Analysis) is a dimensionality reduction technique. LDA is used to maximizes the between – class scattering matrix measure whileminimizes the within – class scatter matrix measure, which make it more reliable for classification [12].

1. Calculate within – class scatter matrix
2. Calculate between-class scatter matrix.
3. Calculate the eigenvectors of the projection matrix W.

2.4 Independent Component Analysis (ICA)

Generalization View of the PCA is known as Independent Component Analysis (ICA) that is used to minimizes the second order and higher order dependencies in the input and determines a set of statistically independent variables [13].The basic steps for ICA algorithm are given below:

1. Collect ô€œºô€¯œ of n dimensional data set X, i = 1, 2, 3, … M.
2. Mean correct all the points: calculate mean and substract it from each data point:.
3. Calculate the covariance matrix :

The ICA of X factorizes the covariance matrix into the following form: where is a diagonal real positive matrix.

F transforms the original data X into Z such that the components of the new data Z are independent: X = FZ [8].

3. LITERATURE REVIEW

FatenBellakhdhar, Kais Loukil and Mohamed ABID“Face recognition approach using Gabor Wavelets, PCA and SVM”2013.In this paper author conveyed that Face recognition is an important research field that is used in pattern recognition. It is necessary for us to give attention to feature extractor and classifier. Because the performance of face recognition system is depend on how to extract feature vector and to classify them into a class correctly. In this paper, the author propose a methodological improvement to raise face recognition rate by fusing the phase and magnitude of Gabor’s representations of the face as a new representation, by using the face recognition algorithm, the principal component Analysis approach and Support Vector Machine

(SVM) that are used in this paper as a new classifier for pattern recognition. The algorithmsthat are proposed in by the author are tested on the public and largely used databases of FRGCv2 face and ORL databases. The algorithm PCA is a global method that is used to implement contrasts with a strong sensitivity to changes in lighting, poses and facial expression (by using the number of poses for each person). The author combinesthe magnitude and the phase of Gabor is used to extract the characteristic vector, the algorithm PCA used for recognition and SVM used to classify faces. Through this approach, it makes the PCA an algorithm effective and commonly used in reducing dimensionality where it can then be used to upstream other algorithms to improve the results of our application[4].

The two operations face extraction and classification are used to improve the performance of face recognition. KiranD.Kadam” Face recognition using Principal Component Analysis with DCT”2014. In this paper the author combining the two techniques of face recognition to increase in recognition rate or increase the accuracy of face recognition system. But there are many methods that work on face recognition. These methods can be divided into three categorization: feature-based matching method, holistic matching method and hybrid methods. The PCA, LDA, ICA methods are belongs to the holistic method which used whole face as input.The two techniques that are used in this paper are Principal Component Analysis (PCA) method which can exactly express every face image via linear operation of Eigen vector and also solve the compression and Discrete Cosine Transform (DCT) method that is used to transform a signal domain in to frequency domain and eliminate the redundancies in an image. The author used the hybrid method of Principal Component Analysis (PCA) and Discrete Cosine Transform (DCT) to reduce the dimension of data on number of images. The complete process of face recognition is shown in the Fig.2.

Figure 2: Algorithm flowchart for face recognition.

To evaluate the performance of these two algorithms PCA and DCT, a code for each algorithm has been generated using matlab. This hybrid technique gives the better result than by using the singlePrincipal Component Analysis (PCA) method. The author achieved the accuracy 99.90% on FACES 94 and 94.70% on ORL. And the recognition rate of single PCA algorithm is less than the combination of PCA and DCT algorithms [5].

PriyankaDhoke, M.P. Parsai”A MATLAB based Face Recognition using PCA with Back Propagation Neural network”2014.The author proposed a method by using Principal Component Analysis (PCA) with Back Propagation Neural Net works (BPNN) method for identification and verification of a person for face recognition system in this paper. The PCA algorithm is used for the reduction of dimensionality of face mage and the recognition is done by the BPNN. The method that is proposed by the author is applied on a database that consists of a set of facial patterns. There are three steps that are used by the author that are the following:-

Figure 3: Face recognition system by using PCA with BPNN.

In the Above Fig. 3, the PCA method is used to reduce dimensionality of an image and each face image may be represented as a feature vector of the eigenfaces, which are stored in a 1D array. By using this feature vector of Eigen faces, the test image can be constructed. The distance between the feature vectors of the test image and that of the database images are then compared. Thus one can reconstruct original image with the help of eigen faces so that it matches the desired image. The back-propagation algorithm is a multi-layer network and it is a fully feed-forward network connection. There are main two layers in this first is input layer and second is output layer. Each layer is connected with another layer and the activation travels from input layer to output layer. Back-propagation algorithm consists of two sweeps of the network which are the forward sweep and the backward sweeps. Forward sweep defines the network from the input layer to the output layer and the backward sweep defines network from the output layer to the input layer.

Figure 4: Diagram of Multilayered neural network.

The author through this paper shows that the PCA for feature extraction and BPNN for image classification and recognition provide the fast computation and high accuracy rate in face recognition system. The execution time of this is only few seconds and acceptance ration s more than 90% [6].

Suman Kumar Bhattacharyya, Kumar Rahul “Face Recognition by Linear Discriminant Analysis” In this paper, the author used the Linear Discriminant Analysis (LDA) method applied to face recognition which is based on a linear projection from the image space to a low dimensional space by maximizing the between class scatter and minimizing the within-class scatter and it the one that method that can maximize the between-class scatter [14]. LDA method overcomes the limitation of Principle Component Analysis method by applying the linear discriminant criterion, in this paper the author take an input test image for identification, the projected test image is compared to each projected training, and the test image is identified as the closest training image and thee image are stored in the ORL face database. And in this paper, the kernel technique is used to project the input data into an implicit space called feature space by nonlinear kernel mapping. The experimental results of this paper show that the correct recognition rate is higher than that of previous techniques by using Discriminant Analysis (LDA) method.

4. CONCLUSION

Through this paper, I will give my opinion that the research work in face recognition is that area in which many people show their interest due to its applications. This paper provides a way which can be understand by all users and readers of all ages in a simplistic, informative and interactive web interface on face recognition. It helps us to solve any problem related to any field for authentication of person by using the given algorithms of Face recognition. There are various methods which exist in the field of face recognition that are used for various fields. From this review paper, I conveyed that the Face recognition system is providing society with new and improved methods of authentication and verification of any person, which makes our work easy in any field.

References

1.en.wikipedia.org/wiki/Facial_recognition_system

2.Preeti S. Subramanyam,Sapana A Fegade”Face and Facial Expression recognition- A Comparative study”2013

3.M.Turk, A.Pentland, “Eigenfaces for Recognition”,Journal of Cognitive Neurosicence, III (1), pp. 71-86, 1991.

4.FatenBellakhdhar, Kais Loukil and Mohamed ABID“Face recognition approach using Gabor Wavelets, PCA and SVM”2013

5.KiranD.Kadam ”Face recognition using Principal Component Analysis with DCT”2014

6.PriyankaDhoke, M.P. Parsai”A MATLAB based Face Recognition using PCA with Back Propagation Neural network”2014

7.Jan Mazanec, Martin Melisek, Milos Oravec, JarmilaPavlovicova “Support Vector Machines, PCA and LDA in face recognition” Journal of Electrical Engineering, VIX(4), pp.203-209,2008.

8.AjeetSingh,BKSingh,ManishVerma” Comparison of Different Algorithms of Face Recognition”2012.

9.http://sebastianraschka.com/Articles/2014_pca_step_by_step.html

10. http://en.wikipedia.org/wiki/Discrete_cosine_transform

11. Aman R. Chadha, Pallavi P. Vaidya, M. Mani Roja” Face Recognition Using Discrete Cosine Transform for Global and Local Features”2011.

12. Yang J., Yu Y. and Kunz W. (2000): An Efficient LDA Algorithm for Face Recognition, the SixthInternational Conference on Control, Automation, Robotics and Vision (ICARCV2000).

13. Bartlett M. S., Movellan J. R., and Sejnowski T. J. (2002): Face Recognition by Independent Component Analysis, IEEE Transactions on Neural Networks, vol. 13, pp. 1450-1464.

14. Mr. Vivek Pali, Mr. Lalit Kumar Bhaiya “Genetic Algorithm Based Feature Selection and BPNN Based Classification for face recognition “Volume 3, Issue 5, May 2013