Continuous Uncertain Interaction

This page accompanies my thesis, "Continuous Uncertain Interaction". PDF versions of the thesis, the complete set of figures used, the thesis BibTeX file, videos and demonstration programs can be found here.

Summary

The thesis examines user interfaces in terms of the of the flow of evidence. The interface is considered as a tool with which users -- in an ongoing, negotiated interaction -- manipulate the beliefs of systems about their intentions. They do this by providing relevant evidence, while the system informs the user as to the effect of evidence upon its beliefs. The construction of the systems which mediate this flow is considered. Predictive displays with appropriate representation of uncertainty are discussed, especially the sonification of changing beliefs. Techniques for bringing long-term inference into tight control loops (and a resulting text entry system) are given, and a new model for continuous, probabilistic selection interfaces is outlined.
Full Abstract

Text

A fully hyperlinked PDF version of thesis is available. Standard Standard version. 247 pages. 40Mb PDF

Figures

High quality extracted figures from the thesis are available for use. See the figures page.

Bibliography

The bibtex file which holds every reference used in the thesis is available here . The references are in the format AbcDefGhiXX, using the first three letters of (up to) the first three author surnames and then the two-digit year. For example, Sheridan and Ferrell 1974 becomes SheFer74.

Videos and Audio

Granular Synthesis and Predictive Displays (Ch. III and IV)

• Video of the mixtures of Gaussians demo. Each labeled target has an associated waveform, and a Gaussian density around its center. As the cursor moves, the timbre smoothly changes. NB: There are some audio glitches due to the screen capturer losing sync.
  WMV
  
  
• Video of the trajectory following demo. The user creates a path, then follows it. Staying on the path, music is timestretched so that the phase is proportional to the arc-length. Leaving the path, background audio is heard.
  WMV
  
  
• Ball-bearing and GPS navigation video, using Monte Carlo sampling to predict possible future actions. (visualisation only).
  WMV
  
  
• The sound generated by the GSLib example code (in Appendix A), showing slowly changing timbre and pitch distributions. WAV
  
  

Hex (Ch. V)

• A video of Hex running on a desktop machine with Monte Carlo prediction enabled.
  WMV
  
  
• Hex running on a PocketPC with accelerometer input.
  AVI
  
  
• Hex running with brain input.
  AVI
  
  

Active Selection (Ch. VI)

• Active selection video for Brownian motion targets. Each individual target moves on a smooth, independent course. The user's mouse actions are applied equally to the trailing targets. Correlating motion results in selection.
  WMV
  
  
• Active selection with the `eggheads' metaphor. A number of `heads' experience orientation disturbances. Input motion is applied to all heads equally. By cancelling the disturbance, selection is achieved.
  WMV
  
  
• Continuous field selection. Selection is performed by holding the desired area of the red curve close to the blue "zero" line. The row of dots at the top give the log probability for that area -- lower indicates higher probability.
  WMV

Demos

Some selected demos are available for download. These demos should not require significant setting up, and should run on most Windows machines. Their purpose and effects are described in more detail in the relevant chapters of the thesis.

• Granular synthesis mixture of Gaussians demo. (Ch. III)
  Instructions: Run granular.exe, with audio enabled. Move the mouse around to hear the changing textures.
  ZIP
  
  
  
• Granular synthesis trajectory demo. (Ch. III)
  Instructions: Run tracepath_thesis.exe. Define a path by left-clicking a number of vertices. Press SPACE when done. Move the mouse along the trajectory, and the music should follow. If you move off, you will have to restart from the beginning of the trajectory.
  ZIP
  
  
  
• Monte Carlo ball bearing demo. (warning, may be unstable on some machines). (Ch. IV)
  Instructions: Start the demo with demo.bat. Use the mouse to change the tilt angle of the surface.
  ZIP
  
  
• Campus exploration demo, with Monte Carlo prediction. (Ch. IV)
  Instructions: Run demo.bat. Use mouse to change heading and prediction horizon. Cursor up and down move forward and backwards. Press 's' to toggle the shadow map. Press 'b' to toggle bike mode. ZIP
  
  
  
• Hex (mouse only version). (Ch. V)
  Instructions: Run hex.exe. Use the mouse to control cursor velocity. Left mouse button autocompletes.
  ZIP
  
  
• Active selection (Brownian motion). (Ch. VI)
  Instructions: Run pctmodel.exe. Use the mouse to move the white targets. All targets will move simultaneously. To select an object, hold the white targetting circle over the the associated yellow target. The object will highlight and eventually will be selected. ZIP
  
  
  
• Active selection with orientation disturbances (Eggheads). (Ch. VI)
  Instructions: Run pctorient.exe. Select an agent by counteracting its natural motion; try and make the head face "forward". The head will light up blue when selection is achieved.
  ZIP
  
  

jhw@dcs.gla.ac.uk
Home