Neural Eye Tracking for Human Computer Interaction

8/2/2019 Neural Eye Tracking for Human Computer Interaction

1/21

Presented By:Sibarama Panigrahi(11040056)VSSUT, Burla


2/21


3/21

Human Computer Interaction(HCI)Literature Review of Eye tracking

How Human eye help in HCI

Why Artificial Neural Network

System Overview

Working of System

Advantage

Limitations Conclusion

Reference

Contents


4/21

Human Computer Interaction

KeyboardTouch

Printer

Mouse

Joystick

Scanner

Speech

BCI

Eye Tracking


5/21

The study of eye movements pre-dates the widespread use of computers byalmost 100 years (Javal, 1878/1879). Beyond mere visual observation,initial methods for tracking the location of eye fixations were quite invasiveinvolving direct mechanical contact with the eye

Dodge and Cline (1901) developed the first precise, non-invasive eyetracking technique, using light reflected from the cornea.

Judd, McAllister & Steel (1905) applied motion picture photography torecord the temporal aspects of eye movements in two dimensions.

Cornsweet and Crane, 1973 uses multiple reflections from the eye couldbe used to dissociate eye rotations from head movement.

Starker & Bolt, 1990 uses eye tracking to serve as a computer inputdevice.

Literature Review


6/21

Curser Control with out mouse

(Much Research is going on)

Using Eye blinking for event handling

(Not yet been much Explored)

How Eye tracking help in HCI


7/21

Why Artificial Neural Network

Head Mounted DevicesSkin ElectrodesContact Lenses withMechanical Devices

Desktop With a CameraUsing ANN

Accurate

Fast

Affordable.

Modest Hardware Requirement

Allows for a fair range of head

movements.


8/21

Eye tracking device/Camera

System Overview


9/21

1. Eye Tracking (i.e. Finding of pupil position)[3][6]

2. Conversion between pupil position & Cursor

Position.[1][2][4]

3. Event handling using Eye Blinking[Proposed]

Working


10/21

1.Eye Tracking [3][6]

Frame Search

Region

Pre-Processing

Bayesian

Classifier

Clustering

Post-Processing

Eyes

Located

Successfully?

Update frameSearch Region

Update Means

And Covariance.

NoLocation Of Eye.

Yes

Frame in


11/21

2.Conversion between pupil position & Cursor Position.

2.1. Data Acquisition Phase (To Train the ArtificialNeural Network(ANN))[2][4]

2.2. ANN For Curser positioning[1][2]


12/21

2.1. Data Acquisition Phase


13/21

2.1. Data Acquisition Phase

The training set is divided into sample Frames of 1/6th of sec.

In each Frame the pupil and its corresponding mouse curser position interms of (x,y) co-ordinates are found and stored as training set.

These mouse curser positions(x,y) are multiplied by a interconnection

weight and passed into first hidden layer.

The number of sample frames depends on duration of data acquisition.

Mathematically nf = t / fsWhere nf=No. of Frames

t=time of data acquisitionfs =Rate at which sample frames are taken

Hence No. of frames in a sec=6


14/21


15/21

2. ANN For Curser positioning

ANN uses Back propagation algorithm for learning where theuser wants to move the curser.

whereXIN and YIN= Final inputs of the neural network.NOX and NOY = Outputs calculated by the neural networksx and y = Coordinates of the eye position

Xscreen and Yscreen =width and the height of the screen, respectively, each in pixels.


16/21

Two eyes blink purposefully to perform double click.

Left eye blinking can be used for left click of mouse.

Right click of mouse can be used for right click of

mouse.

Event handling using Eye Blinking[Proposed]


17/21

Eye tracking can allow people to use their eyes to communicateleading to Faster Interaction.

Modest hardware requirement.

Reduced cost.

Its users include people withBrain injuries

Cerebral palsy

Multiple sclerosis

Muscular dystrophy

Spinal cord injuriesSpinal muscular atrophy

Advantages


18/21


19/21

Although eye tracking for human ComputerInteraction is a technology over the last 20 yearsbut still a lot of aspects like eye blinking can beused for handling events , head position movement,etc will help in designing more interactiveinterfaces leading to ambient intelligence.

Conclusion


20/21

1. E. DEMJN1, V. ABOSI2, Z. TOMORI1 , Eye Tracking Using Artificial NeuralNetworks for Human Computer Interaction Physiol. Res. 60: 841-844, 2011 .

2. Anaelis Sesin, PhD; Malek Adjouadi, PhD;* Mercedes Cabrerizo, PhD; MelvinAyala, PhD; Armando Barreto, PhD, Adaptive eye-gaze tracking using neural-network-based user profiles to assist people with motor disability, Journal ofRehabilitation Research & Development Volume 45,Pages 801818 , Number 6,2008.

3. A. Haro, M. Flicker and I. Essa, Detection and tracking eyes by using theirphysiological properties, dynamics and appearance, Proceedings of IEEE CVPR2002.

4. Adjouadi M, Sesin A, Ayala M, Cabrerizo M. Remote eye gaze tracking systemas a computer interface for persons with severe motor disability. In: Lecturenotes in computer science. Vol. 3118. Berlin (Germany): Springer; 2004. p.76169.

5. Robert J.K. Jacob, Ph.D., Keith S. Karn, Ph.D. Commentary on Section 4. Eyetracking in human-computer interaction and usability research: Ready to deliverthe promises.

6. Subramanya Amarnag, Raghunandan S. Kumaran and John N. Gowdy, REALTIME EYE TRACKING FOR HUMAN COMPUTER INTERFACES Proceedings of IEEE CVPR 2006.

References


21/21

Neural Eye Tracking for Human Computer Interaction

Documents

Transcript of Neural Eye Tracking for Human Computer Interaction