Face recognization using artificial nerual network

Face Recognition using Artificial Neural Network

Presented by

Dharmesh R Tank(13014081024)

M Tech – CE (Sem III)

Guided by

Assist Prof D S Pandya

Prof Menka Patel

Outline

Objective

History

Basic Concept

Proposed FC System

Discrete Cosine Transform

Artificial Neural Network with Back Propagation

Thresholding Rule

Applications

References

Objective

Face recognition, most relevant applications of image analysis.

True challenge to build an automated system which equals human ability to recognize faces.

Humans are quite good identifying known faces, but not very skilled when large amount of unknown faces.

Human face recognition ability help to develop a non-human face recognition system.

History

Engineering started to show interest in facerecognition in the 1960’s. One of the firstresearches on this subject was Woodrow W.Bledsoe.

In 1960, Bledsoe, along other researches, startedPanoramic Research, Inc., in Palo Alto, California.

The majority of the work is AI-related contractsfrom the U.S. Department of Defense andvarious intelligence agencies.

A simple search with the phrase “FaceRecognition” in the IEEE Digital Library throws9422 results. 1332 articles in only one year -2009.

Basic Concept

Some face coordinates were selected by a humanoperator, and then computers used this information forrecognition.

Face recognition is used for two primary tasks:

Verification (one-to-one matching)

Identification (one-to-many matching)

Even 50 years later Face Recognition still suffers -variations in illumination, head rotation, facialexpression, aging, occlusion.

Still new problems to measure subjective face featuresas ear size or between-eye distance are on thecontinuity basis.

Face Detection Feature Extraction Face Recognition

Problems with

Existing

High information redundancy

Maintain a huge database of faces

Computationally expensive

Energy compaction issues

Occlusion, face rotation, illumination, facial expression, aging

Proposed Face

Recognition System

Input Images

Face Detection

Feature Extraction(DCT)

Normalization & Classification

(ANN)

Face Recognition

Output

Discrete Cosine

Transform

Basis functions for N = 8

DCT[2] is applied to the entire face image to obtain all frequency components of the face.

DCT[3] is used as a tool for dimensionality reduction to extract illumination invariant features.

Image is said to be DC free, after removing first DCT coefficient.

Remove the redundant information

Decrease the computational

complexity(orthogonal)

Much faster than any other models

(Linear)

Energy compact

Example[5]

Discrete Cosine

Transform

The DCT is defined as:

The Inverse DCT is defined as:

Where

Artificial Neural

Network

ANN[1] are computational models inspired byan animal's central nervous systems (inparticular the brain) which is capableof machine learning as well as patternrecognition.

Artificial neural networks are generallypresented as systems of interconnected"neurons" which can compute values frominputs.

Adaptive Learning

Self Organization

Self Classification

ANN Architecture

I[7]

Σ

f

Output Y

Input X1,X2,X3......Xn

Weights (W1,W2,W3……..Wn)

Fig 1.1 ANN Procedure

ANN Architecture

II

Input Layer Output Layer

Hidden Layer

Fig 1.2 Two layer Artificial Neural Network

Back Propagation

[10]

Trains the network to achieve a balance between the ability to respond correctly to the input patterns that are used for training.

Ability to provide good response to the input that are similar.

Requires a dataset of the desired output for many input, making up the training set.

Method calculates the gradient of a loss function with respects to all the weights in the network.

The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the loss function.

These are necessarily Multilayer Perceptron[11](MLPs).

Multilayer Perceptron

(MLP) Neural

Network

It is a three layers architecture. Input for NN is a grayscale image.

Number of input units is equal to the number of pixels in the image.

Number of hidden units.

Number of output unit is equal to the number of persons to be recognized.

Every output unit is associated with one person.

NN is trained to respond “+1” on output unit, corresponding to recognized person.

For other aliens images output will be “-1” . We called this perfect output.

Thresholding Rule

Introduce thresholding rules, which allow improving recognition performance by considering all outputs of NN.

Known as ‘square rule’.

Calculates the euclidean distance between perfect and real output for recognized person.

When this distance is greater than the threshold, rejection take place. Otherwise acceptation.

The best threshold is chosen experimentally.

Literature Review[2]

Rising Year What we get

1950 Human Psychology Studies

1960 Born of Face Recognition field by Woodrow W. Bledsoe at Panoramic Research

1964-65 Bledsoe, along with Helen Chan and Charles Bisson, worked on using computers to recognize human faces

1971 Bell Laboratories by A. Jay Goldstein, Leon D. Harmon and Ann B. Lesk, vector, containing 21 subjective features like ear protrusion, eyebrow weight or nose length, as the basis to recognize faces using pattern classification techniques

1973 Fischler and Elschanger tried to measure similar features automatically

1973 Kenade, developed a fully automated face recognition system. Kenade compares this automated extraction toa human or manual extraction, showing only a small difference. He got a correct identification rate of 45-75%.

Continues…

Rising Year What we get

1980 Mark Nixon, presented a geometric measurement for eye spacing . This decade also Some researchers build face recognition algorithms using artificial neural networks.

1986 Eigenfaces in image processing, a technique thatwould become the dominant approach in following years, was made by L. Sirovich and M. Kirby

1992 Mathew Turk and Alex Pentland of the MIT presented a work which used eigenfaces for recognitionPCA(Principal Component Analysis), ICA(Independent Component Analysis), LDA(Linear Discriminant Analysis)

Applications

Areas Applications

Information Security Access Security / Data Privacy / Authentication

Access Management Access Log / Permission Based System

Biometrics Person Identification (Passports, Voter ID, Driver licenses) / Automated identity verification (border controls)

Law Enforcement Video Surveillance / Suspect Identity / Suspect Tracking / Simulated Aging

Personal Security Home Video Surveillance Systems / Expression Interpretation (Driver Monitoring System)

Entertainment Leisure Home Video Game / Photo Camera Applications

Real Time Application

Microsoft’s Project Natal[12]

Toyota are developing sleep detectors to increase safety[13]

Sony’s PlayStation Eye[14]

Google Glass with DNN[16]

References

[1] Three approaches for face recognition V.V. Starovoitov1, D.I Samal1,D.V. Briliuk1, The 6-th International Conference on Pattern Recognitionand Image Analysis October 21-26, 2002, Velikiy Novgorod, Russia, pp.707-711

[2] Face Recognition Algorithms, Proyecto Fin de Carrera, June 16, 2010

[3] A Literature Survey on Face Recognition Techniques, Riddhi Patel#1,Shruti B.Yagnik, IJCTT) – volume 5 number 4 –Nov 2013

[4] Face Recognition Using Artificial Neural Network , A. E. Shivdas Deptof E & T Engineering, RIT, Maharashtra, India, IJRMST (E-ISSN: 2321-3264) Vol. 2, No. 1, April 2014

[5] High Speed Face Recognition Based on Discrete Cosine Transformsand Neural Networks.ppt

[6] High Speed Face Recognition System Based on DCT and RBF NNMeng Joo Er, Weilong Chen, and Shiqian Wu IEEE Transactions onNeural Network Volume 16, Number 3, May 2005

[7] A Introduction to Natural Computation, Lecture 08, Perceptrons byLeandro Minku

References

[8] http://en.wikipedia.org/wiki/Artificial_neural_network

[9] http://www.slideshare.net/ArtificialNeuralNetwork

[10] http://en.wikipedia.org/wiki/Backpropagation

[11] http://en.wikipedia.org/wiki/Multilayer_perceptron

[12] B. Dudley. ”e3: New info on microsoft’s natal – how it works, multiplayer and pc versions”. The Seattle Times, June 3 2009.

[13] K. Massy. ”toyota develops eyelid-monitoring system”. Cnetreviews, January 22 2008.

[14] M. McWhertor. ”sony spills more ps3 motion controllerdetails to devs”. Kotaku. Gawker Media., June 19 2009.

[15]http://kotaku.com/5297265/sony-spills-more-ps3-motion-controllerdetails-to-devs.

[16] www.nametag.ws

[17] http://www.kdnuggets.com/2014/06/new-beginnings-facial-recognition.html

Thank You Question ??