Thursday, January 21, 2010

Noise tolerant selection by gaze-controlled pan and zoom in 3D

Authors:
Dan Witzner Hansen
Henrik H. T. Skovsgaard
John Paulin Hansen
Emilie Mollenbach

Summary:
Stargazer - a circular keyboard with a gaze typing process. 3 Concentric circles are used as cursor showing the point of interest and direction of gaze. Auditory feedback is used. Zoom and pan technique is used to navigate through the keyboard. Saccadic movements - fast movements by the user to get the overview of the selectable objects. Zoom - gazing at a particular area makes it possible to disregard the area of least interest. Pan - Horizontal and vertical translation in the same scale of zoom. Allows smooth pursuit movements to seek the object of intent. Dwell time activation - object is selected if the user stares at the object for certain time. The stargazer was tested small, medium and large sized displays. It was evaluated based on speed (words per minute), error rate ( number of backspace / number of characters) and the remaining errors (number of manipulations required to produce a string). WPM were significantly high for large and medium displays than the small display. The remaining errors remained to low. This states that the navigation scheme used was easy to understand. Users could cope with latency upto 200 ms. Users with 5 minutes of training could perform the typing faster.

Discussion:
Clustered objects on the screen slows down the speed of selection. It would be interesting to see if restricting the number of letters on the keyboard based on the dictionary removes clutter and increase the selection speed.
can there by fatigue while gazing? Will it affect the speed of selection ?

Comments:

5 comments:

  1. Yes, that is something else I was thinking - at some point, the user might get pretty tired of looking in certain patterns.

    ReplyDelete
  2. I don't remember if the paper said anything about clutching, but it seems that that would get to be more of a problem as fatigue increases.

    ReplyDelete
  3. I'm pretty sure that users will tire from using gaze-based interaction too much, since it's unnatural for people to force their eyes to make such movements for that long of a period of time.

    Also, I believe less choices usually implies less clutter, but at a cost of an extra step, but that might be a topic for another paper.

    ReplyDelete
  4. Reducing the number of targets presented may not be necessary in this particular example as with practice one should be able to make semi automatic saccads as the location of the letters are initially constant, analogous to typing on a keyboard.

    ReplyDelete
  5. I think showing only those keys which a dictionary will allow in a given state is a good idea to make the interface less cluttered, but it will also constraint the users in being able to type only those words that are in a dictionary. Sometimes the user might want to have full control. What we will have to think about then is how to switch back and forth between the two modes seamlessly.

    ReplyDelete