ISSN 2278-6856 A NEW APPROACH OF FACE RECOGNITION · PDF filethis technique is low computational requirement, high ... PCA face recognition algorithm with ... for low cost hardware

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS) Web Site: www.ijettcs.org Email: [email protected], [email protected]

Volume 3, Issue 3, May-June 2014 ISSN 2278-6856

Volume 3, Issue 3 May – June 2014 Page 238

Abstract: Face recognition is the most effective and natural technique to identify a person. This paper introduces a new technique to recognize human face artificially using DCT, PCA and SOM neural network. Principal component analysis (PCA) is a classical and successful method of dimension reduction. Discrete Cosine Transform (DCT) is a well known compression technique and Self Organize Map (SOM) act as a classifier and has been used for face space representation. The evaluation of the work is done in MATLAB by taking 16 images of 4 different people and each person has 4 diverse facial expressions. The main advantage of this technique is low computational requirement, high speed and better recognition rate. Keywords: Discrete Cosine Transform, Principal Component Analysis, Self Organize Map. 1. INTRODUCTION Face recognition is a challenging and interesting research topic in the field of pattern recognition which has been found a widely used in many applications such as verification of credit card, security access control, and human computer interface. Thus many face recognition algorithms have been proposed and survey in this area can be found in [2] [3] [4]. Face recognition procedure works in three stages: Face Detection, Feature Extraction and Face classification or recognition. The problem of face recognition is complicated by age, skin, color and gender also with differing image qualities, facial expressions, facial furniture, background, and illumination conditions [1]. Among various solutions to the problems of face recognition [5] the most successful seems to be those appearance-based approaches, which generally operate directly on images or appearances of face objects and process the image as two dimensional patterns. The main trend in feature extraction has been representing the data in a lower dimensional space computed through a linear or non-linear transformation satisfying certain properties. Principal component analysis (PCA) [12],[13], linear discriminate analysis (LDA) [6] and discrete cosine transform (DCT) [7],[8],[9],[10] are three main techniques used for data reduction and feature extraction in the appearance-based approaches. This paper presents a new approach to recognize face using DCT, PCA and self organize map as classifier.

2. OVERVIEW OF PROPOSED WORK In this paper a new improved and fast approach for face recognition is considered which relevant to increase the recognition rate and to reduce the training time as compared to previous researches. The work integrates the PCA face recognition algorithm with well known DCT compression algorithm to reduce the amount of time to finish up the recognition of the face. The algorithm has been tested on 4 subjects of 4 different facial expressions. A face image of the untrained data will be inputted to the system, the system will apply first the DCT for the compression of the images and then feature extraction technique will be applied by the PCA algorithm. Finally the SOM neural network is used to classify the data and to compare with the database whether it is found in the database or not. The advantages of using DCT and PCA and SOM are described next in this section. A block diagram of the proposed work of face recognition is shown in the following figure.

Figure 1: Block diagram of the proposed work

3. OBJECTIVES Main objectives of the proposed work are as follows: (a) To design a model for an ideal facial recognition system. (b) To enhance the model for a high-speed facial Recognition system (c) To develop a program in MATLAB based on the designed model. (d) To evaluate a face recognition training time and recognition rate by combining 2D-DCT and PCA with neural network. 4. METHODOLOGIES Methodologies used in Proposed work: DCT, PCA, SOM are described as (A) 2-D Discrete Cosine Transform: In [14], Compression is an important process by which the description of computerized information is modified so that the capacity required to store or the bit-rate required to transmit it is reduced. The DCT is a transform which transforms a signal or image from the spatial domain to the elementary frequency domain. Lower frequencies are

A NEW APPROACH OF FACE RECOGNITION USING DCT, PCA, AND

NEURAL NETWORK IN MATLAB

1Urvashi Bakshi 2Rohit Singhal

1 Research Scholar, Institute of Engineering and Technology, Alwar, Rajasthan, India 2 Associate Professor, Institute of Engineering and Technology, Alwar, Rajasthan, India




more obvious in an image than higher frequencies an image is transferred into its frequency components and higher frequency coefficients are discarded, the amount of data needed to describe the image without sacrificing too much image quality will reduce. (B) Principal Component Analysis : In [15], PCA is a useful statistical technique that has found application in fields such as face recognition and image compression, and is a common technique for finding patterns in data of high dimension. It is a way of identifying patterns in data, and expressing the data in such a way as to highlight their similarities and differences. The other main advantage of PCA is that once you have found these patterns in the data, and you compress the data, i.e. by reducing the number of dimensions, without much loss of information. (C) Self Organizing Map: In [16], the principal goal of self-organizing maps is to transform an incoming signal pattern of arbitrary dimension into a one or two-dimensional discrete map, and to perform this transformation adaptively in a topologically ordered fashion. Self-organizing maps learn to recognize groups of similar input vectors in such a way that neurons physically near each other in the neuron layer respond to similar input vectors.

Figure 2: Self-Organising-Map

5. EXPERIMENTAL WORK The experimental work of the proposed system consist the following procedural steps: A. Image Database: First step is the collection of all the images to create a database for the face recognition technology. The database is divided into two subsets, for training and testing purposes. During SOM training, 16 images were used, containing four subjects and each subject having 4 images with different facial expressions. Figure 2 shows the training and testing image database constructed.

Figure 3: Training database

Figure 4: Testing Database: Untrained face images with

different envrironment and background

B. Pre-processing: All face images are preprocessed in adobe photoshop to give them a fix size of 180 by 200. Then import all faces images into MATLAB workspace and convert it into grayscale image. C. Compress all face images using DCT: DCT compression technique is then applied to all the images, to test the recognition system the technique is performed on all 4 test images. D. Extract features using PCA: Principal Component Analysis was then applied to all the images for dimension reduction and for extracting features of one and all face images. E. Reshaping the image size: the pre-processed gray scale images of 180by200 sizes are reshaped in MATLAB having size 8 × 8 pixels into a 64 × 1 array with 64 rows and 1 column for each image. Similarly, the image database for training uses these images and forms a matrix of 64 × 20 with 64 rows and 20 columns. F. Design a SOM neural network: To represent the intensity level of gray scale pixels the input vectors defined for the SOM are distributed over a 2D-input space varying over [0 255]. G. Give training to neural network and simulate the images: The input vectors used to train the SOM with dimensions [64 2], where 64 minimum and 64 maximum values of the pixel intensities are represented for each image sample. The SOM created with these parameters is a single-layer feed forward SOM map with a competitive transfer function and 128 weights. The weight function of this network is the negative of the Euclidean distance [5]. These images are used for test the network and without any overlapping the training and testing set were used.

Figure 5: Euclidean Distance Weight for Trained Image

Database with SOM Network.

H. Result Analysis of training time and recognition rate:




Figure 6: Result of face recognition system in same and

different environment-closest match is found The result obtained from this simulation shows that a facial image having different facial expression can be easily identified. After SOM training, the 20-dimensional trained image database is transformed into a 64-dimensional map where the magnitude of the layer weights is increased and Euclidean distance for feature vector for trained image is smaller than untrained image. This transformation produces better classification by grouping similar clusters together. This experiment concern with the processing time of overall system and this is the time that required for training the SOM network. Main goal of this experiment is to reduce the training time while maintaining the recognition rate. Table 1 shows the recognition rate and training time that is achieved in the case of 100 epochs. Table 1: Effects of epochs on accuracy or recognition rate No. of epochs

No. of images

Training time(sec)

Recognition rate

100 16 2.58 97.5 500 16 12.22 97.5 1000 16 21.2 97.5 6. CONCLUSION AND FUTURE WORK This paper has presented a novel face recognition technique that uses DCT for compression, PCA as dimensionality reduction and SOM as a classifier. The system was evaluated in MATLAB using an image database of 16 face images having 4 subjects and each subject having 4 images with different facial expressions and different backgrounds. After training for approximately 1000 epochs the system achieved better recognition rate and training time with lesser epochs. The

system is also having less computational requirement this make system well suited for low cost hardware implementation. Instead of PCA, LDA can also be used to achieve good results, we hope to implement the fusion of DCT with LDA and SOM. REFERENCES [1] J. Nagi, “Design of an Efficient High-speed Face

Recognition System”, Department of Electrical and Electronics Engineering, College of Engineering, University Tenaga National, March 2007.

[2] W. Zhao, R. Chellappa, A. Rosenfeld and P. J. Phillips, “Face Recognition: A Literature Survey”, ACM Computing Surveys, vol. 35, No. 4, 2003, pp.399 - 458.

[3] Ashok Samal and Prasana A.Iyengar, “Automatic recognition and analysis of human faces and facial expressions: A survey”, Pattern Recognition, vol. 25, 1992, pp.65-77.

[4] R. Chellappa, C.L. Wilson, and S. Sirohey, “Human and machine recognition of faces: A survey”, Proceedings of the IEEE, vol. 83, 1995, pp.705-740.

[5] J. R. Solar, P. Navarreto, " Eigen space-based face recognition: a comparative study of different approaches”, IEEE Tran. , Systems man And Cybernetics- part c: Applications, Vol. 35, No. 3, 2005.

[6] M. Turk, A. Pentland, "Eigen faces for face recognition", Journal cognitive neuroscience, Vol. 3, No.1, 1991.

[7] W. Zhao, R. Chellappa, A, Krishnaswamy, “Discriminant analysis of principal component for face recognition”, .IEEE Trans. Pattern Anal. Machine Intel., Vol 8, 1997.

[8] M.J.Er, W.Chen, S.Wu, “High speed face recognition based on discrete cosine transform and RBF neural network”, IEEE Trans. On Neural Network, Vol. 16, No. 3, PP. 679,691, 2005.

[9] D. Ramaeubramanian, y. Venkatesh, “Encoding and recognition of Faces based on human visual model and DCT”, Pattern recognition, Vol. 34, PP. 2447-2458, 2001.

[10] X. Y. Jing, D. Zhang, “A face and palm print recognition approach based on discriminant DCT feature extraction”, IEEE trans. on Sys. Man & Cyb., Vol. 34, No. 6, PP. 2405-2415, 2004.”

[11] D. Kumar, C.S. Rai, and S. Kumar, “Face Recognition using Self- Organizing Map and Principal Component Analysis” in Proc. on Neural Networks and Brain, ICNNB 2005, Vol. 3, Oct 2005, pp. 1469-1473.

[12] D.L. Swets and J.J. Weng, “Using Discriminant Eigen features for image retrieval”, IEEE Trans. Pattern Anal. Machine Intel, vol. 18, PP. 831-836, Aug. 1996.

[13] P.N. Belhumeur, J.P. Hespanha, and D. J. Kriegman, “Eigen faces vs. Fisher faces: Recognition using class




specific linear projection”, IEEE Trans. Pattern Anal. Machine Intel., vol. 19, PP. 711-720, may 1997.

[14] Andrew B. Watson. “Image Comp ression using the Discrete Cosine Transform”, NASA Ames Research Center, Florida, United States of America. Mathematica Journal, 4(1), 1994. p. 81- 88.

[15] A.H.Boualleg, Ch.Bencheriet and H.Tebbikh, 2006, “Automatic Face Recognition Using Neural Network- PCA”, Proc. 2nd Information and Communication Technologies ICTTA '06, pp. 1920 – 1925, Damascus, Syria.

[16] “Self-Organizing Maps. Wikip edia Source”, http://en.wikipedia.org/wiki/Self-organizing map, Retrieved 6 January 2014

Authors:

Ms. Urvashi Bakshi received her B.tech degree in IT in 2008, She is a research scholar in IT department in I.E.T college Alwar, Rajasthan, and doing further research on Face

Recognition Technology.

Mr. Rohit Singhal is a B.tech and M.tech in Computer Science and Engineering Stream. He is a Associate Professor in CSE department in I.E.T. college, Alwar,

Rajasthan, his area of interest is Image Processing and Computer Graphics.

ISSN 2278-6856 A NEW APPROACH OF FACE RECOGNITION · PDF filethis technique is low computational requirement, high ... PCA face recognition algorithm with ... for low cost hardware

Documents