Tovi Grossman, Ravin Balakrishnan

Dep. of Computer Science Univ. of Toronto

CHI 2004

Introduction

Background

Goals and Directions of the Current Study

Experiment

Implications for User Interface Design

Conclusion

Advances in three dimensional display technology

Appropriate user interfaces

: easily select and manipulate virtual elements in the 3D display space.

They studied and modeled user performance in the most fundamental

interaction task – pointing – in a 3D display.

Contents of paper

review previous work on pointing and Fitts’ law models in 1D and 2D

identify various factors and interactions between them in 3D pointing

propose several mathematical models

controlled experiment

Implications for user interface design

Fitts’ law One dimension

Two dimension (Mackenzie and Buxton, 1992)

Accot and Zhai (2003) : A weighted Euclidean model

Ware et al. (1994, 1997) : 3D model, similar to 2D model

A

W

A

W

H

Experimental Hardware Platform : volumetric display was used Manipulation of Experimental Parameters

Target dimension : W, H, D Movement angle : Ө uses a approach angle of 0°

Modeling First baseline model & Weighted model

Weighted Euclidian model

Incorporating f( ) into the IDӨ Wtmin model

Incorporating f( ) into the IDӨ WtEuc model

Apparatus

Participants 5 female and 7 male, ages from 20 to 25

Procedure : cuboids target, XZ plane, 3D cursor Design

A(3)ⅩH(4)ⅩW(4)ⅩD(4)Ⅹangle(3) = 576 combinations, 3 sessions, random order 6 reciprocal movements, 2 practices, 6 groups, 1 hour per each session

Performance measure Movement time, number of errors per trial

Movement time analysis Main effect : all had a main effect on MT (angle : 0, 45, 90) Effect of movement direction : no effect Interactions ( Ө↔ W, D, H) : no HⅩθ → targets were located in the XZ plane Relative effect of target dimensions : Figure 4e Interactions betw. target dimensions : Figure 5

Movement time analysis Main effect : all had a main effect on MT (angle : 0, 45, 90) Effect of movement direction : no effect Interactions ( , W, D, H) : Ө no HⅩθ Relative effect of target dimensions Interactions betw. target dimensions

Movement time analysis Fit of the models

Error analysis : effects for D,H,W, no effect for A, error rate 15.7%, 0.04 units → error ↑

“racing through the experiment”

Target sizing

target size → error rate

dimension parallel to the line of approach

Target positioning

moving forward and backward → left and right

adjust the layout of the interface elements according to the position of users

A true 3D volumetric display

Moving forwards and backwards in depth is

slower than moving left and right for selecting targets.

The target width was more critical than the height and depth of the target.

The effect of the height was constant regardless of the movement angle.

Accounts for varying movement angles

The implications of our results for the sizing and positioning of user

interface widgets in 3D displays

Accot, J., & Zhai, S. , Refining Fitts' law models for bivariate pointing. ACM

CHI. p. 193-200., 2003

MacKenzie, S. , Fitts' law as a research and design tool in human-computer

interaction, Human-Computer Interaction. 7. p. 91-139., 1992

MacKenzie, S., & Buxton, W., Extending Fitts’ law to two-dimensional

tasks, ACM CHI. p. 219-226., 1992

※ Murata, A., Iwase, H., Extending Fitts’ law to a three-dimensional pointing

task, Human Movement Science Vol 20, pp.791-805., 2001