COMMENT
1.
2.
SUMMARY
This paper describes HoloSketch, a virtual reality-based 3D geometry creation and manipulation tool. It uses a head-tracked stereo shutter glasses and a desktop CRT display. The system allows users to create virtual objects (and view the results in real time stereoscopically) with a 3D wand manipulator directly in front of the user, at very high accuracy. For every frame rendered for each eye a viewing matrix is calculated using the information from head tracker. The authors claim that the system is very accurate (compared to existing VR technologies at that time) with an error of +-0.25cm in world coordinates.
Menu Design: It is a fade up menu which appears when the right wand button is pressed. The menu is organized as a 3D pie menu and appears to be right in front of the wand pointer. To select user pokes the want tip at the desired button. The menu also supports sub menus. A sub menu appears in the front pushing back the main menu in space.
Features:
1. HoloSketch supports many drawing primitives like rectangular solids, spheres, ellipsoids etc. Other types of drawings involve 'sweeping' across the space and putting drawing material. The drawing material can be wireframe lines, toothpaste etc. In addition to creating new geometry, new geometries can be imported from other modelling environments.
2. Color selection can be done using a color cube.
3. An object can be selected by pressing the middle want button over it. A rubber band selection is also available for group selection.
4. Editing operations like translation, scaling, grouping of objects, rotation, animation are supported.
5. Viewing Control: The viewer can look around objects by moving his head. The viewer can also move/rotate the entire virtual universe. The viewer can also move through the virtual universe.
Results: To test their system he authors engaged an artist (having no prior experience with holosketch but experience with tradition 3d modelling systems) for a month of trails using their system. They found that it was very quick and easy to create complex forms using holosketch. They also fund that it was difficult to make fine adjustments holding ones hands in air. To remedy this the 10X reduction mode was used. In future author would like to expand the study to test for novice users. The author also plans to improve and extend the animation controls.
DISCUSSION
This seems to be a seminal paper in stereoscopic 3d modelling given that it was published in 1995. The menu design seems to be intuitive and useful for selecting in 3D. The idea that we can look around objects and move in the 3D virtual universe seems very interesting. I would have loved to see a user study involving novice users and their feedback. The authors plan to do that in future.
Tuesday, January 26, 2010
Wearable EOG goggles: eye-based interaction in everyday environments
COMMENT
1.
2.
SUMMARY
The paper presents an embedded eye tracker which relies on Electroculography (EOG) for tracking. The eyes are the origin of a steady electric potential field. Movement of eyes from the center towards the periphery results in a change in this potential. Hence conversely, if we study the change in this potential, position of the eye ball can be determined. The system described in this paper consists of goggles with the following devices connected:-
1. 4 electrodes are arranged near the left eye. These two pairs of electrodes sense the change in potential due to the movement of the eye and send the signals to the EOG processor.
2. A fifth electrode mounted above the right eye provides the reference signal.
3. A pocket worn component for EOG signal processing produces the horizontal and vertical movement of the eyes which can be transmitted to a computer using bluetooth.
Eye gesture recognition: Following steps are involved in the recognition-
1. Blink detection and removal: Blinks are detected using a template matching approach using a template created manually from example blinks from different persons.
2. Saccade detection: This is done using the Continuous Wavelet transform.
3. Eye gesture recognition: Each gesture is encoded in basic directions first which are L, R, U and D (left, right, up, down). In the second step diagonal eye movements are combined which are represented as 1, 3, 7 and 9. To recognize gestures this gesture string is then compared with known templates.
To evaluate the EOG system, a game with 8 different levels was used. Subjects had to repeatedly perform eye gestures as accurately and as fast as possible. To reach a high score wrong eye movements had to me minimized. The authors found that EOG is a robust modality of HCI applications. However 30% of the subjects reported getting fatigued.
DISCUSSION
Producing unnatural gestures with eyes seems to me a not very good mode of interaction. The visual channel is already overloaded with many tasks - gazing on the screen, typing, reading information. Adding another voluntary task would result in fatigue and confusion in my opinion. Also I do not see the application for the kind of gestures described in the paper. The authors have not motivated their paper in that area. The user study was lacking in that it does not say much about how easy users found to perform gestures, how accurate the system was etc.
1.
2.
SUMMARY
The paper presents an embedded eye tracker which relies on Electroculography (EOG) for tracking. The eyes are the origin of a steady electric potential field. Movement of eyes from the center towards the periphery results in a change in this potential. Hence conversely, if we study the change in this potential, position of the eye ball can be determined. The system described in this paper consists of goggles with the following devices connected:-
1. 4 electrodes are arranged near the left eye. These two pairs of electrodes sense the change in potential due to the movement of the eye and send the signals to the EOG processor.
2. A fifth electrode mounted above the right eye provides the reference signal.
3. A pocket worn component for EOG signal processing produces the horizontal and vertical movement of the eyes which can be transmitted to a computer using bluetooth.
Eye gesture recognition: Following steps are involved in the recognition-
1. Blink detection and removal: Blinks are detected using a template matching approach using a template created manually from example blinks from different persons.
2. Saccade detection: This is done using the Continuous Wavelet transform.
3. Eye gesture recognition: Each gesture is encoded in basic directions first which are L, R, U and D (left, right, up, down). In the second step diagonal eye movements are combined which are represented as 1, 3, 7 and 9. To recognize gestures this gesture string is then compared with known templates.
To evaluate the EOG system, a game with 8 different levels was used. Subjects had to repeatedly perform eye gestures as accurately and as fast as possible. To reach a high score wrong eye movements had to me minimized. The authors found that EOG is a robust modality of HCI applications. However 30% of the subjects reported getting fatigued.
DISCUSSION
Producing unnatural gestures with eyes seems to me a not very good mode of interaction. The visual channel is already overloaded with many tasks - gazing on the screen, typing, reading information. Adding another voluntary task would result in fatigue and confusion in my opinion. Also I do not see the application for the kind of gestures described in the paper. The authors have not motivated their paper in that area. The user study was lacking in that it does not say much about how easy users found to perform gestures, how accurate the system was etc.
Distant freehand pointing and clicking on very large, high resolution displays
COMMENT
1. Comment on Drew's blog
SUMMARY
The paper discusses system for pointing and clicking interaction with very high resolution displays from a distance. The pointing interaction enabled by this system is absolute (as if the user is pointing with a laser beam emitted from his hand) as opposed to pointing using a mouse where position of the mouse and the position of the cursor on the screen are not direct mapped. Clicking and grabbing are performed by using hand gestures. Prototype:- User wears reflective markers on the hand and fingers. The large display used is rendered using a cluster of 18 PCs. The paper explores two clicking gestures - AirTap (tapping with the index finger in air) and Thumb trigger (moving the thumb in and then outwards from the index finger). However users found the thumb trigger tiring and the authors adopted a style similar to air tap as the style of interaction. The authors also explored 3 pointing techniques:-
1. Ray casting. The cursor is placed at a point where the ray coming out from the tip of the intersects with the plane of the screen. Air tap method of clicking could not be used with this method due to the induced jitter.
2. Relative pointing and clutching: In this method the open hand is used for relative cursor control. Clenched hand is used to disengage the hand from the pointer and recalibrate it.
3. Hybrid technique:- Hybrid method uses ray casting for recalibration, whereas cursor movement is done via relative hand movement.
In their study, there was no significant difference between air tab and thumb trigger. Ray casting as a method of pointing resulted in high error rates and uncomfortable interaction. The evaluation was conducted with 12 participants. Users were asked to perform two tasks - Transition task and sequence task. Ray casting was found to be generally faster but also very error prone. There was no significant difference between Ray and Ray to relative. participants however, preferred 'relative' because the clutching action was similar to lifting a mouse (on which they were already trained).
DISCUSSION
Overall the paper provides an in depth study of different interaction techniques with large displays. Regarding the jitter induced by air tap and the its claimed incompatibility with Ray casting, I feel a data glove would be able to solve this problem. I feel that when users are pointing using ray casting the movements involved are in the forearm and the shoulder, while the joints in the hand remain fixed. However, during clicking the shoulder and the forearm remain fixed whereas the joints on the index finger move. So side effect Air tap clicking could be removed by ignoring any changes to cursor position when only the joints in the hand are moved.
1. Comment on Drew's blog
SUMMARY
The paper discusses system for pointing and clicking interaction with very high resolution displays from a distance. The pointing interaction enabled by this system is absolute (as if the user is pointing with a laser beam emitted from his hand) as opposed to pointing using a mouse where position of the mouse and the position of the cursor on the screen are not direct mapped. Clicking and grabbing are performed by using hand gestures. Prototype:- User wears reflective markers on the hand and fingers. The large display used is rendered using a cluster of 18 PCs. The paper explores two clicking gestures - AirTap (tapping with the index finger in air) and Thumb trigger (moving the thumb in and then outwards from the index finger). However users found the thumb trigger tiring and the authors adopted a style similar to air tap as the style of interaction. The authors also explored 3 pointing techniques:-
1. Ray casting. The cursor is placed at a point where the ray coming out from the tip of the intersects with the plane of the screen. Air tap method of clicking could not be used with this method due to the induced jitter.
2. Relative pointing and clutching: In this method the open hand is used for relative cursor control. Clenched hand is used to disengage the hand from the pointer and recalibrate it.
3. Hybrid technique:- Hybrid method uses ray casting for recalibration, whereas cursor movement is done via relative hand movement.
In their study, there was no significant difference between air tab and thumb trigger. Ray casting as a method of pointing resulted in high error rates and uncomfortable interaction. The evaluation was conducted with 12 participants. Users were asked to perform two tasks - Transition task and sequence task. Ray casting was found to be generally faster but also very error prone. There was no significant difference between Ray and Ray to relative. participants however, preferred 'relative' because the clutching action was similar to lifting a mouse (on which they were already trained).
DISCUSSION
Overall the paper provides an in depth study of different interaction techniques with large displays. Regarding the jitter induced by air tap and the its claimed incompatibility with Ray casting, I feel a data glove would be able to solve this problem. I feel that when users are pointing using ray casting the movements involved are in the forearm and the shoulder, while the joints in the hand remain fixed. However, during clicking the shoulder and the forearm remain fixed whereas the joints on the index finger move. So side effect Air tap clicking could be removed by ignoring any changes to cursor position when only the joints in the hand are moved.
Noise tolerant selection by gaze-controlled pan and zoom in 3D
COMMENT
1. Comment on Manoj's blog
SUMMARY
This paper presents StarGazer, a 3D interface for gaze-based
interaction and target selection using continuous pan and zoom. One of the goals of the paper is to work with cheap low resolution eye trackers. The authors mention that dwell time has been the common way of identifying gaze based selection. However, they argue that this method is unnatural and significantly reduces the speed of the user. The paper presents and zoom and pan based method to to perform this interaction. The format of StarGazer is generic but the authors have described it in context gaze based typing. They describe a circular keyboard with letters arranged in concentric circles. The zooming process reduces the area under contention and hence reduces noise. To navigate within the zoomed in view the user can use panning. In the beginning of a selection user makes a fast saccadic movement, since the confidence here is low, zoom is not used. Once the gaze settles on a selectable object, the object is panned to the center. If the user continues to focus, the object enters the zoomed area and when it gets sufficeintly close (defined by a plane of activation), it is selected. Audio and visual (color of selected object changes) feedback is given. Studies were conducted to see if the Star Gazer interface was easy to use for novice users. 48 users performed a typing task. The mean error rate for large displays was 5.6%. Another test was conducted to measure the typing speed of the users using this interface. Subjects produced 8.16 words per minute with a mean error rate of 1.23%.
DISCUSSION
Although the paper presents a novel way to interact with gaze, I feel that, as is true with all gaze based interfaces, this will lead to considerable fatigue. The paper does not describe the perception of subjects who used the interface. I would have liked to see some qualitative evaluation in addition to the indepth quantitative one which they have provided.
1. Comment on Manoj's blog
SUMMARY
This paper presents StarGazer, a 3D interface for gaze-based
interaction and target selection using continuous pan and zoom. One of the goals of the paper is to work with cheap low resolution eye trackers. The authors mention that dwell time has been the common way of identifying gaze based selection. However, they argue that this method is unnatural and significantly reduces the speed of the user. The paper presents and zoom and pan based method to to perform this interaction. The format of StarGazer is generic but the authors have described it in context gaze based typing. They describe a circular keyboard with letters arranged in concentric circles. The zooming process reduces the area under contention and hence reduces noise. To navigate within the zoomed in view the user can use panning. In the beginning of a selection user makes a fast saccadic movement, since the confidence here is low, zoom is not used. Once the gaze settles on a selectable object, the object is panned to the center. If the user continues to focus, the object enters the zoomed area and when it gets sufficeintly close (defined by a plane of activation), it is selected. Audio and visual (color of selected object changes) feedback is given. Studies were conducted to see if the Star Gazer interface was easy to use for novice users. 48 users performed a typing task. The mean error rate for large displays was 5.6%. Another test was conducted to measure the typing speed of the users using this interface. Subjects produced 8.16 words per minute with a mean error rate of 1.23%.
DISCUSSION
Although the paper presents a novel way to interact with gaze, I feel that, as is true with all gaze based interfaces, this will lead to considerable fatigue. The paper does not describe the perception of subjects who used the interface. I would have liked to see some qualitative evaluation in addition to the indepth quantitative one which they have provided.
Introduction
- What is your e-mail address (e.g., yourname at domain.com)? akb2810 at gmail dot com.
- What is your academic standing? 2nd year Masters, Computer Science
- Where are you originally from? Any one heard of Kashmir, India. I am from there.
- Why are you taking this class? 1. Interested in the field of HCI. 2. Liked the special topics course which I took previously with Dr. Hammond.
- What do you expect to be doing in 10 years? Be an expert in some field. It could be on a job or something I am doing on my own.
- What do you think will be the next biggest technological advancement in computer science? Computers as we know them getting obsolete. We will just have terminals (popularly called thin clients) and all computation and storage will move to the cloud.
- What current sight and/or touch-related technology do you find interesting and why? Except for the IPhone and Microsoft Surface computer, I haven't come across any such technologies in the market. IPhone was very interesting few years ago, now its behavior is more expected than interesting. I like Microsoft's vision for Surface Computing .. well I forgot what it exactly was.. but something to the effect that every surface should come alive - the walls, table tops and what have you.
- If you could have lunch with someone famous in history, who would it be and why? After thinking a lot about this I decide to go beyond history. It would be Krisna, the hero from the Indian epic Mahabharata. If he indeed wrote the book called Gita, I would like to discuss rationality with him over lunch.
- What is your favorite movie? Hotel Rwanda, and one Indian movie called Swades.
- Tell us some details about yourself that you would like to add to the class poster. Every once in a few years I paint. I think about it daily.
Subscribe to:
Posts (Atom)
