Cs--eye Gaze For Reding Detection
This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA
Overview
Download & View Cs--eye Gaze For Reding Detection as PDF for free.
More details
- Words: 1,100
- Pages: 16
National Institute of Science & Technology
Eye Gaze for Reading Detection
EYE GAZE FOR READING DETECTION PRESENTED BY:
ASWINI KUMAR PANDA Roll # CS200117324 UNDER THE GUIDANCE OF
Mr. R. K. Shial Aswini Kumar Panda
[1]
National Institute of Science & Technology
Eye Gaze for Reading Detection
INTRODUCTION •
There has been an increasing interest in the fields of Human Computer Interfaces (HCI).
•
Here we examine the possibility of determining whether a computer user is engaged in reading material, which is displayed on the computer monitor, by analyzing video sequences of the user.
•
Gaze movement data has led to two very different types of interfaces: command and non-command.
•
In the typical window interface, there are five primary goals to eye movements: (a) inspecting, (b) searching, (c) exploring, (d) performance monitoring, and (e) reading.
Aswini Kumar Panda
[2]
National Institute of Science & Technology
Eye Gaze for Reading Detection
UTILITY OF DETECTING READING •
In windows-based operating systems, eye-gaze is a valuable way, where a fixation threshold determines interest in some display object.
•
Reading detection provides a much more precise means for determining user interest because it can determine the level of user interest based on the type of user behavior like reading,scanning or skimming.
•
implementing our reading detection method is to obtain information about what text the user is reading in order to infer user interests and then adapt information displays to user needs.
Aswini Kumar Panda
[3]
National Institute of Science & Technology
Eye Gaze for Reading Detection
UTILITY OF DETECTING READING contd.. •
More accurate feedback results in more accurate models of the user.
•
Another specific advantage is that by using gaze movement data, computer help systems could be given more context information and therefore provides more accurate help.
•
Finally, knowledge of whether the user is reading, skimming, or scanning would be useful for creating adaptive peripheral displays.
Aswini Kumar Panda
[4]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM •
It processes raw video to infer reading activity.The first major step is to estimate gaze direction; these are then used for inferring reading.
•
To obtain gaze direction, detecting skin colored regions and then detecting faces in these regions carries out face localization. Next, eye localization is performed within the detected face. Once localized they are tracked.
•
One localized or tracked eye in each frame is then used to determine the person’s gaze direction. The gaze information is then used for reading detection.
Aswini Kumar Panda
[5]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
•
Changes that take place in the gaze directions over a specified time interval during reading activity are modeled using a set of Finite State Machines (FSMs).
•
Gaze directions estimated over a finite time interval (window) are checked with respect to these FSMs to determine whether the interval represents reading.
Aswini Kumar Panda
[6]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
1. Determining gaze direction •
A common method employed in the process of face localization is skin colour segmentation and for this reason a face detector is used .
•
The next step of the design involves eye localization which can be done 2 methods:a. One that uses colour information of iris and sclera. b. Blink detection.
Second method is better. Aswini Kumar Panda
[7]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
2. Proposed algorithm for blink detection. •
Blink detection allows localization of the eyes, by using frame differencing followed by optical flow.
•
Subsequently, the localized eyes are tracked by the KanadeTomasi-Lucas method. i. Steps in Blink detection ii. Tracking of the Eyes iii. Speed Considerations
Aswini Kumar Panda
[8]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
•
Detecting when a user is reading rather than merely scanning or skimming from eye-gaze patterns is a difficult problem.
•
Our reading detection system relies on three mechanisms: (a) coarse or quantized represented of eye-movements, (b) pooled evidence based detection, and (c) mode switching.
•
Two methods are generally involved:
Aswini Kumar Panda
[9]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
I. First method: •
The system is initially in scanning mode, which requires a set of events to occur to switch into reading mode. The events that are tracked include the specific eye movements shown in Table .
•
The quantized, tokenized stream of eye-movement data is then pooled to determine whether the user is reading.
•
The pooled evidence for reading is calculated by taking the accumulated value of the pooled data and adding the points associated with the current event for both the X and Y axes.
Aswini Kumar Panda
[10]
National Institute of Science & Technology
Eye Gaze for Reading Detection
Table:
Aswini Kumar Panda
[11]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
II.FSM method: •
FSMs are well suited to model eye gaze transitions and they are computationally inexpensive.
•
The inputs to the reading detection algorithm from the gaze direction module are either “LEFT” or “RIGHT” for every frame.
•
Two pieces of information are extracted from the periodic movements of the gaze during reading.
•
An additional variable, TCounter, is used to keep track of the amount of time the FSM stays in each state.
Aswini Kumar Panda
[12]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
An example of FSM with 4 states
Aswini Kumar Panda
[13]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM •
CONTD.
After constructing the database of FSMs representing reading, reading detection is done in 2 steps. 1. The first step involves building a FSM for the presented gaze directions over an interval and comparing it with those in the database. 2. Reading detections over a set of 6 consecutive windows is performed and the majority vote of these 6 detections is used to determine whether reading is taking place.
Aswini Kumar Panda
[14]
National Institute of Science & Technology
Eye Gaze for Reading Detection
CONCLUSION Here the algorithm proposed to differentiate reading activities from other non-reading ones, such as playing computer games and watching video clips, on a computer monitor. An average reading detection accuracy of 85.4% on 10 people was obtained with a false alarm rate of 14.2%.
Aswini Kumar Panda
[15]
National Institute of Science & Technology
Eye Gaze for Reading Detection
Thank You…
Aswini Kumar Panda
[16]
Eye Gaze for Reading Detection
EYE GAZE FOR READING DETECTION PRESENTED BY:
ASWINI KUMAR PANDA Roll # CS200117324 UNDER THE GUIDANCE OF
Mr. R. K. Shial Aswini Kumar Panda
[1]
National Institute of Science & Technology
Eye Gaze for Reading Detection
INTRODUCTION •
There has been an increasing interest in the fields of Human Computer Interfaces (HCI).
•
Here we examine the possibility of determining whether a computer user is engaged in reading material, which is displayed on the computer monitor, by analyzing video sequences of the user.
•
Gaze movement data has led to two very different types of interfaces: command and non-command.
•
In the typical window interface, there are five primary goals to eye movements: (a) inspecting, (b) searching, (c) exploring, (d) performance monitoring, and (e) reading.
Aswini Kumar Panda
[2]
National Institute of Science & Technology
Eye Gaze for Reading Detection
UTILITY OF DETECTING READING •
In windows-based operating systems, eye-gaze is a valuable way, where a fixation threshold determines interest in some display object.
•
Reading detection provides a much more precise means for determining user interest because it can determine the level of user interest based on the type of user behavior like reading,scanning or skimming.
•
implementing our reading detection method is to obtain information about what text the user is reading in order to infer user interests and then adapt information displays to user needs.
Aswini Kumar Panda
[3]
National Institute of Science & Technology
Eye Gaze for Reading Detection
UTILITY OF DETECTING READING contd.. •
More accurate feedback results in more accurate models of the user.
•
Another specific advantage is that by using gaze movement data, computer help systems could be given more context information and therefore provides more accurate help.
•
Finally, knowledge of whether the user is reading, skimming, or scanning would be useful for creating adaptive peripheral displays.
Aswini Kumar Panda
[4]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM •
It processes raw video to infer reading activity.The first major step is to estimate gaze direction; these are then used for inferring reading.
•
To obtain gaze direction, detecting skin colored regions and then detecting faces in these regions carries out face localization. Next, eye localization is performed within the detected face. Once localized they are tracked.
•
One localized or tracked eye in each frame is then used to determine the person’s gaze direction. The gaze information is then used for reading detection.
Aswini Kumar Panda
[5]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
•
Changes that take place in the gaze directions over a specified time interval during reading activity are modeled using a set of Finite State Machines (FSMs).
•
Gaze directions estimated over a finite time interval (window) are checked with respect to these FSMs to determine whether the interval represents reading.
Aswini Kumar Panda
[6]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
1. Determining gaze direction •
A common method employed in the process of face localization is skin colour segmentation and for this reason a face detector is used .
•
The next step of the design involves eye localization which can be done 2 methods:a. One that uses colour information of iris and sclera. b. Blink detection.
Second method is better. Aswini Kumar Panda
[7]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
2. Proposed algorithm for blink detection. •
Blink detection allows localization of the eyes, by using frame differencing followed by optical flow.
•
Subsequently, the localized eyes are tracked by the KanadeTomasi-Lucas method. i. Steps in Blink detection ii. Tracking of the Eyes iii. Speed Considerations
Aswini Kumar Panda
[8]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
•
Detecting when a user is reading rather than merely scanning or skimming from eye-gaze patterns is a difficult problem.
•
Our reading detection system relies on three mechanisms: (a) coarse or quantized represented of eye-movements, (b) pooled evidence based detection, and (c) mode switching.
•
Two methods are generally involved:
Aswini Kumar Panda
[9]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
I. First method: •
The system is initially in scanning mode, which requires a set of events to occur to switch into reading mode. The events that are tracked include the specific eye movements shown in Table .
•
The quantized, tokenized stream of eye-movement data is then pooled to determine whether the user is reading.
•
The pooled evidence for reading is calculated by taking the accumulated value of the pooled data and adding the points associated with the current event for both the X and Y axes.
Aswini Kumar Panda
[10]
National Institute of Science & Technology
Eye Gaze for Reading Detection
Table:
Aswini Kumar Panda
[11]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
II.FSM method: •
FSMs are well suited to model eye gaze transitions and they are computationally inexpensive.
•
The inputs to the reading detection algorithm from the gaze direction module are either “LEFT” or “RIGHT” for every frame.
•
Two pieces of information are extracted from the periodic movements of the gaze during reading.
•
An additional variable, TCounter, is used to keep track of the amount of time the FSM stays in each state.
Aswini Kumar Panda
[12]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM
CONTD.
An example of FSM with 4 states
Aswini Kumar Panda
[13]
National Institute of Science & Technology
Eye Gaze for Reading Detection
ALGORITHM •
CONTD.
After constructing the database of FSMs representing reading, reading detection is done in 2 steps. 1. The first step involves building a FSM for the presented gaze directions over an interval and comparing it with those in the database. 2. Reading detections over a set of 6 consecutive windows is performed and the majority vote of these 6 detections is used to determine whether reading is taking place.
Aswini Kumar Panda
[14]
National Institute of Science & Technology
Eye Gaze for Reading Detection
CONCLUSION Here the algorithm proposed to differentiate reading activities from other non-reading ones, such as playing computer games and watching video clips, on a computer monitor. An average reading detection accuracy of 85.4% on 10 people was obtained with a false alarm rate of 14.2%.
Aswini Kumar Panda
[15]
National Institute of Science & Technology
Eye Gaze for Reading Detection
Thank You…
Aswini Kumar Panda
[16]
Related Documents
Cs--eye Gaze For Reding Detection
June 2020 3
The Gaze
July 2020 11
Biological Systems For Intrusion Detection
May 2020 7
Data Mining For Intrusion Detection
October 2019 33
Twenty Tips For Lowering The Gaze-munazid
May 2020 8
Gaze Vapori Iritanti
April 2020 4More Documents from "Dr Diana Enachescu"
Technical Seminar Presentation 2004
June 2020 4
En Neurocomp
June 2020 7
Artificial Neural Network For Misuse Detection
June 2020 8