IRJET- Deep Feature Fusion for Iris Biometrics on Mobile Devices
0 likes25 views
This document proposes a deep neural network-based system for iris recognition and authentication on mobile devices. The system uses feature fusion with multiple feature extraction algorithms to recognize iris patterns. It then uses a backpropagation neural network for classification and authentication. The proposed system aims to more accurately recognize iris patterns, reduce processing time, and provide appropriate authentication outputs compared to existing CNN-based systems. It segments the iris from input images, extracts features using algorithms like FAST, SURF, and Harris, and feeds these into a neural network for classification and authentication. Performance charts are generated to evaluate the system. The proposed system seeks to securely authenticate users for mobile applications and transactions.
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 2469
The above figure is a block diagram for classification
technique where neural network is used. The pre-processed
output and segmented output (feature extracted output
value) is given as test pattern to the neural network. It
compares the result of test pattern with the stored train
pattern. Thus the classification of pattern is done and the
authentication is granted when weights of test and train
pattern are equal. The access is denied if the pattern weights
are unequal.
6.5 Neural network chart
A neural network chart is finally generated in order tocheck
the performance, training state, error histogram, regression
and fit of the process. The validation check, time,
performance, gradient and epoch (number of iterations)are
shown in this chart. A dialog box which shows the
authentication result is shown after the generation of the
performance chart. The dialog box contains a grant message
when the iris pattern matches with the user’s pattern and it
contains denied message when the iris patterns does not
matches.
Fig -6: Performance Chart
6.6 Output
Fig -7: Dialog box
The output of the system is a dialog box that contains
“ACCESS GRANTED” or “ACCESS DENIED” message.
7. METHODOLOGY
The method showcased here is the pattern recognition and
authentication system. This system uses iris image of user’s
iris as input. It is pre-processed and segmented. Then
the pattern is recognized by feature extraction
algorithm. The neural network uses back propagation
algorithm for classifying the pattern and determines
authentication. The performance chart for the process
is generated to check the accuracy. The output is a
dialog box with an access granted or denied message.
8. CONCLUSION
The proposed system has the ability to recognize the
accurate iris pattern of user and grant authentication of
different applications or accounts. There are many
advantages with the proposed system as compared with the
existing system which include low cost, low processingtime,
appropriate output and accurate authentication. By this
model we can easily secure our personal or financial
accounts and other application and need not to memorize
passwords.
REFERENCES
[1]S. Thavalengal and P. Corcoran, “User authentication on
smartphones: focusing on iris biometrics,” IEEE Consumer
Electronics Magazine, vol. 5, no. 2, pp. 87–93, 2016.
[2] J. G. Daugman, “High confidence visual recognition of
persons by a
test of statistical independence,” IEEE Transactions on
Pattern Analysis and Machine Intelligence,vol.15,no.11,pp.
1148–1161, 1993.
[3] K. B. Raja, R. Raghavendra, and C. Busch, “Smartphone
based robust iris recognition in visible spectrum using
clustered k-means features,” inIEEEWorkshoponBiometric
Measurements and Systems for Security and Medical
Applications, 2014, pp. 15–21.](https://image.slidesharecdn.com/irjet-v6i3479-190827063522/85/IRJET-Deep-Feature-Fusion-for-Iris-Biometrics-on-Mobile-Devices-3-320.jpg)
IRJET- Deep Feature Fusion for Iris Biometrics on Mobile Devices
- 1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 2467 Deep Feature Fusion for Iris Biometrics on Mobile Devices Abinaya M1, Aarthika R2, Prof Rajat Kumar Dwibedi3 1,2Students, Dept. of Electronics and Communication Engineering, Jeppiaar SRR Engineering College, Chennai, Tamil Nadu 4Assistant Professor, Dept. of Electronics and Communication Engineering, Jeppiaar SRR Engineering College, Chennai, Tamil Nadu ---------------------------------------------------------------------***---------------------------------------------------------------------- Abstract – If the fast growing mobile technology, when it comes to sensitive transactions such as financial or payment applications, it is required to follow the security in all kinds of transactions made through mobile devices. Image based biometric authentication creates good impact on security. In the existing system a deep feature fusion network is used that exploits the unique information in iris regions. It first applies max-out units into the convolution neural network (CNN) to generate a compact representation foreach modalityandthen fuses the difference in the features of two modalitiesthrougha weighted concatenation. In the proposedsystemadeepneural network based classification algorithm is used, edgedetection with different feature extraction is used for recognition. The proposed system authenticate for a particular access and which can be implemented in MATLAB software. Key Words: Biometric, Deep Neural Network, Feature fusion, MAT lab software, Authentication. 1. INTRODUCTION In today’s world, hacking of personal and financial account on mobile devices is a biggest problem. Authentication with face, fingerprint and voice plays a prominent role. Iris biometric recognition and authentication is more secure compared to other biometric authentication. In the existing systems the methodology for recognition is not appropriate. The proposed system uses back propagation algorithm for comparison. For recognition of pattern feature fusion with different feature extraction algorithm is used. The proposed system consists of three modules which are pre-processing module, segmentation module and classifier module. Thus the output is more appropriate and gives accurate authentication. 2. OBJECTIVE The main objective is to recognise the authorised user’s iris pattern and grant authentication for access on mobile devices. This model proposes an efficient and effective way for iris recognition and authentication using back propagation algorithm which is a deep neural network technique. The pattern recognition is done using different feature extraction algorithm so that non users can’t use the personal or financial accountswithouttheuser’spermission. 3. EXISTING SYSTEM In the existing system convolutional neural network is used for comparison of pattern which is limited to orientation identification, processing time anddatastoragecapacity.The output is not appropriate with this technique. Moreover the pattern of user iris is recognized with a single extraction algorithm. This islimitedtocertainenvironmentalconditions so that the pattern is not recognized accurately and the authentication is inappropriate. The quality of image is problem in this method. This system is not secure as it gives authentication even to non-users. 4. PROPOSED SYSTEM The proposed system overcomes the limitations of the existing system. The input imageispre-processedinorder to enhance the features of the image. The pattern is recognised by feature fusion where different feature extraction algorithms are used and the value of each extraction algorithm is considered and the pattern is chosen, so the pattern is more accurate. A back propagation algorithm which is comprised under deep neural network is used in order to compare the iris patterns. The proposed system gives a chart that indicatestheperformance,errorhistogram training state, processing time and validation checks. This makes the output more appropriate, the processing time is minimised and so the proposed system is effective and efficient. 5. BLOCK DIAGRAM The fig 1 shows the block diagram of the proposed system. Fig -1: Block diagram
- 2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 2468 6. SYSTEM DESIGN 6.1 Neural Network Fig -2: Neural Network Neural Networks are organisation of layers which are made up of multiple nodes interconnected with functions. Neural network reads the relationship between input and output data layers with the help of the hidden layers.Thisprocessis done through finding the connection in weights of input layer with hidden layer as well as hidden layer with output layer. Neural Networks derives the meaning of complicated and imprecise data. The process of neural network issimilar to the function of neurons in human brain. The different set of data is analysed by Neural Network and they are trained to find the connections between different data and give corresponding outputs. They are used for pattern recognition, adaptive learning, real-time operation and self- organisation. 6.2 Pre-Processing Module Fig -3 Pre-Processing flow chart The pre-processing is done to enhance the features of input image. The input image is the user’s eye image which converted from RGB to GRAY image. The binary conversion takes place as these values are required to train the Neural Network. The edge detection is used for finding the boundary of the iris image. SOBEL edge detection is used as it is more suitable for finding the edges of image without noise and its values. At last resizing of image is done to get same size of image each and every time. 6.3 Segmentation and Feature Extraction Fig – 4: Feature Extraction In this module, segmentation of iris and feature extraction for pattern recognition is done. Segmentation is partitioning of image. Here iris is segmented from the input image. Different feature extraction algorithms are carriedouttoget the values of feature points. FAST, SURF, HARRIS, MIN EIGEN, MSER feature extraction algorithms are used. The unique pattern of user’s iris image is found through this process. The determined values are given as input to the neural network. 6.4 Classification Fig -5: Authentication process
- 3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 2469 The above figure is a block diagram for classification technique where neural network is used. The pre-processed output and segmented output (feature extracted output value) is given as test pattern to the neural network. It compares the result of test pattern with the stored train pattern. Thus the classification of pattern is done and the authentication is granted when weights of test and train pattern are equal. The access is denied if the pattern weights are unequal. 6.5 Neural network chart A neural network chart is finally generated in order tocheck the performance, training state, error histogram, regression and fit of the process. The validation check, time, performance, gradient and epoch (number of iterations)are shown in this chart. A dialog box which shows the authentication result is shown after the generation of the performance chart. The dialog box contains a grant message when the iris pattern matches with the user’s pattern and it contains denied message when the iris patterns does not matches. Fig -6: Performance Chart 6.6 Output Fig -7: Dialog box The output of the system is a dialog box that contains “ACCESS GRANTED” or “ACCESS DENIED” message. 7. METHODOLOGY The method showcased here is the pattern recognition and authentication system. This system uses iris image of user’s iris as input. It is pre-processed and segmented. Then the pattern is recognized by feature extraction algorithm. The neural network uses back propagation algorithm for classifying the pattern and determines authentication. The performance chart for the process is generated to check the accuracy. The output is a dialog box with an access granted or denied message. 8. CONCLUSION The proposed system has the ability to recognize the accurate iris pattern of user and grant authentication of different applications or accounts. There are many advantages with the proposed system as compared with the existing system which include low cost, low processingtime, appropriate output and accurate authentication. By this model we can easily secure our personal or financial accounts and other application and need not to memorize passwords. REFERENCES [1]S. Thavalengal and P. Corcoran, “User authentication on smartphones: focusing on iris biometrics,” IEEE Consumer Electronics Magazine, vol. 5, no. 2, pp. 87–93, 2016. [2] J. G. Daugman, “High confidence visual recognition of persons by a test of statistical independence,” IEEE Transactions on Pattern Analysis and Machine Intelligence,vol.15,no.11,pp. 1148–1161, 1993. [3] K. B. Raja, R. Raghavendra, and C. Busch, “Smartphone based robust iris recognition in visible spectrum using clustered k-means features,” inIEEEWorkshoponBiometric Measurements and Systems for Security and Medical Applications, 2014, pp. 15–21.
- 4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 2470 BIOGRAPHIES Abinaya M Pursuing DegreeinElectronicsand Communication Engineering in Jeppiaar SRR Engineering College, Chennai, Tamil Nadu. Aarthika R Pursuing DegreeinElectronicsand Communication Engineering in Jeppiaar SRR Engineering College, Chennai, Tamil Nadu. Mr Rajat Kumar Dwibedi M.Tech Assistant Professor in Electronics andCommunicationEngineeringin Jeppiaar SRR Engineering College, Chennai, Tamil Nadu.