Authentication system using EEG biometric for smart home

Jian-feng Hu1, a, Zhen-dong Mu 2,b

1 Institute of Information Technology, Jiangxi University of Technology, Nanchang 330098, China

2Institute of Information Technology, Jiangxi University of Technology, Nanchang 330098, China

ahuguess@qq.com,b418623577@qq.com

Keywords: EEG, Biometric, smart home, authentication system

Abstracts: smart home, is a trend of home furnishing development, smart home operation should

not be too complicated, it will lose the meaning of smart home, the EEG signals, different from the

degree of response to familiar objects, using the EEG signal successfully solve the smart home in

the room into the authentication problem.

Introduction

Smart living is envisioned to be the standard of living for many people in near future. Smart living

enables easy and understanding style of living where all computing is silently, seamlessly embodied

in many human’s life aspects. One smart living dream that attracts many researcher interests is

smart home. Researchers aspire to have intelligent smart home that is capable to cater daily human

needs in automatic fashion. Many researchers are geared towards realizing the function. However,

there are still few researchers focus on security aspect of entry access related to smart home. Smart

home authentication system is an urgent system to be installed within smart home’s compounds.

Conventional authentication systems use secret knowledge like password. They are deemed

inconvenience and difficult to remembering passwords and easy-to-crack passwords of secret

words. Biometric appears to answer the problem related to conventional system.

One of the ways to implement biometric authentication system is by authenticating them via

image or video captured using a dedicated terminal as biometric enrollment module. This biometric

module is expensive and can be destroyed by thieves to bypass biometrics authentication after alarm

system being turn off.

I this paper, we present our study on an approach to authenticate smart home users through

their EEG conveniently. Our proposed approach is very easy and simple thereby its simplicity

allows very fast extraction.

Related work on smart home authentication

EEG characteristics: In the intelligent home furnishing authentication, the authentication and

the biggest difference is that the authentication intelligent home furnishing is finally verified by

people can be determined, so the key point in the intelligent home furnishing authentication is to

prevent the non through the use of. We use the mature visual evoked potential of EEG is more

appropriate, because compared to visual evoked potential and electrical signals to other brain,

relatively stable feature, the study also more mature, for use in the detection process, in recent years

the use of visual evoked potential of personal identification has some achievements, such as the

study group the use of photo identification studies, reveals the different user to own photo reaction

Applied Mechanics and Materials Vols. 457-458 (2014) pp 1228-1231Online available since 2013/Oct/31 at www.scientific.net© (2014) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/AMM.457-458.1228

All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of TTP,www.ttp.net. (ID: 130.194.20.173, Monash University Library, Clayton, Australia-06/10/14,14:00:36)

induced by users, with characteristics of EEG signals, thus realizing the identification. In the

intelligent home furnishing in the system, can use and users or items of interest, such as music

stimuli evoked tools, the following figure shows the photo when authenticating users EEG features.

ms-100.0 150.0 400.0 650.0 900.0

uV 0.0

-5.0

-10.0

-15.0

-20.0

-25.0

5.0

10.0

15.0

20.0

25.0

*ms12-22.avg

ms3-22.avg

ms22-22.avg

Subject: EEG file: ms12-22.avg Recorded : 11:14:41 28-Dec-2012Rate - 1000 Hz, HPF - 0 Hz, LPF - 300 Hz, Notch - 50 Hz

NeuroscanSCAN 4.3Printed : 15:32:43 24-Jun-2013

Fig1 Pay attention to their own and others comparison chart

Data captured：Acquisition device users with EEG, adjust the relevant parameters, according to

the instruction, look at the screen display, the EEG acquisition can be divided into two parts, one for

the user the EEG feature extraction and classifier to establish part, this part is to set up user EEG

identification model and related parameters, real-time acquisition of EEG and the second is the

implementation of intelligent home furnishing users.

Extraction: The original EEG after the above method interception, converting the time domain

signal into frequency domain by AR model, according to the different subjects characteristics are

extracted, and was simulated with BP neural network, obtained for different subjects. In this paper,

we use Hjort derivation to reduce interference from the neighboring electrode.

The Hjort derivation H

iC is calculated as

∑∈

−=iSj

ji

H

i sccC4

1

(1)

Where ci is the reading of the center electrode scj, with i=1…30 and j is the set of indices

corresponding to the eight electrodes surrounding electrode ci.

EEG signal acquisition band 0.05Hz ~ 200Hz, in order to extract features, is first filtered EEG,

acquisition of band 0.05Hz ~ 50Hz.

Time-domain EEG data disorganized EEG in order to better highlight the characteristics of

EEG signal, we use AR model to convert the time domain signals into frequency domain, and

extract the feature from the frequency domain signals.

The Fisher distance was often used to denote differences between classes in classification

research. The bigger the fisher distance was the more notable the difference was.

The Fisher class reparability criterion was used preparatory to extract features. The Fisher

distance of two classes was calculated as

2

2

2

1

2

21 )(

σσµµ

+−=F (2)

Where µ was equalizing value and σ was variance.

Applied Mechanics and Materials Vols. 457-458 1229

Multilayer back-propagation neural networks were trained to classify any two of the four

classes’ motor imagery EEG. The two classes were coded by 1 output unit. The hidden layer

consisted of 20 units. The input layer had 200 units, 100 representing a channel and 100 for another.

The traingdx learning algorithm has been used to train the network, which uses gradient

descend with momentum and variable learning rate in batch learning mode. Gradient descend with

momentum can avoid a shallow local minimum and a variable learning rate can make the learning

as fast as possible while maintaining stability. The batch learning was used to update the network

weights after all training data was presented.

Simulation and experimental results

In order to test whether can use EEG control, we conducted a simple experiment, the experiment

is designed to test the intelligent home furnishing room in the system, whether the owner can enter the

room. The theory is based on when the subjects saw familiar items, reflecting the goods more than

other people to see, we according to the theory of testing a group of experimenters, a group of

experimenters consists of five people, each test thirty times, the recognition system recognition rate

as shown in fig:

1 2 3 4 50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Fig 2. Correctly identify host ratio

1 2 3 4 50

0.2

0.4

0.6

0.8

1

Fig 3. Correctly identify stranger ratio

In order to better results, we design a set of other people, the same group of people is five, and the

purpose of the experiment is to identify refused entry to the test system, the results as shown in figure

3. From the above two results, a good system to identify the owner, and good refused to strangers.

1230 Frontiers of Mechanical Engineering and Materials Engineering II

Conclusion and Future work

We present our study about a novel intelligent system for smart home. In EEG biometric

details, we take the EEG recognition idea. We present an approach to improve EEG recognition

towards noise. We believe this idea will result on efficient and convenient authentication system

when deployed in smart home.

Acknowledgment

This work was supported by Natural Science Foundation of Jiangxi Province [No

20132BBE50051]

Reference

[1]Palaniappan R, Mandic D P. EEG Based Biometric Framework for Automatic Identity

Verification. The Journal of VLSI Signal Processing, 2007, 49(2): 243-250.

[2]Palaniappan R. Method of identifying individuals using VEP signals and neural network. IEE

Proceedings - Science, Measurement and Technology, 2004, 151(1): 16-20.

[3]Palaniappan R. Electroencephalogram signals from imagined activities: a novel biometric

identifier for a small population. Intelligent Data Engineering and Automated Learning

(IDEAL), Lecture Notes in Computer Science 2006, 42: 604-611.

[4]Touyama H, Hirose M. Non-target photo images in oddball paradigm improve EEG-based

personal identification rates. Annual International Conference of the IEEE Engineering in

Medicine and Biology Society, 2008, 1:4118-21.

Applied Mechanics and Materials Vols. 457-458 1231

Frontiers of Mechanical Engineering and Materials Engineering II 10.4028/www.scientific.net/AMM.457-458 Authentication System Using EEG Biometric for Smart Home 10.4028/www.scientific.net/AMM.457-458.1228

DOI References

[2] Palaniappan R. Method of identifying individuals using VEP signals and neural network. IEE Proceedings

- Science, Measurement and Technology, 2004, 151(1): 16-20.

http://dx.doi.org/10.1049/ip-smt:20040003