visual analytics methodology for eye movement studieszwang/schedule/dc11.pdfvisual analytics...

16
Visual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel Weiskopf 04/10/2013 Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel Weiskopf KSU Graphic & Visualization Lab Seminar

Upload: others

Post on 30-May-2020

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analytics Methodologyfor Eye Movement Studies

Gennady Andrienko, Natalia Andrienko, Michael Burch, DanielWeiskopf

04/10/2013

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 2: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Summary

Eye movement recordings are viewed as a window into internalcognitive process (the ”eye-mind” hyptothesis)

HCI and visualization researchers hope to understand user’sinformation processing and factors affecting the usability ofthe displays and interfaces

Visual anslysis of eye tracking data with Geo-VA methods

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 3: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

IntroductionEye tracking data

Record format: user identifier, time, position in the displayspace and fixation duration.

Types of eye movement analysis tasks

Tasks focusing on areas of interest (AOIs)Focusing on movements.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 4: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresList of the analysis methods

MT: map display of trajectories.STC: space-time cube display of trajectories.PSA: path similarity analysis consisting of computation ofpairwise distances between trajectories, projection andgropuing of the trajectories by similarity.FM: flow map of summarized eye movement.AM: summary map of spatial distribution of users’ attention.CTF: clustering of time intervals by similarity of the spatialpatterns of flows.CTA: clustering of tim eintervals by similarity of the spatialpatterns of attention distribution.TVT: temporal view of trajectories showing attributes oftrajectory segments, such as the distance to a selected AOI.TSF: filtering of trajectory segments.TEE: extraction of events from trajectories.FSD: discovery of frequent sequences of area visits.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 5: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresData Transformations

Adjustment of Time References:aligning the start and/or end times of multiple trajectories.

Spatial Generalization:replacement of the original spatial positions in the trajectoriesby coarse space units/areas.

Spatio-Temporal Aggregationtrajectories are transfromed into sequences consisting of visitsof the places and moves (transitions) between them.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 6: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresExploring and Comparing Individual Trajectories

A: a map display of multiple trajectories shown with 20%opacity.B,C: map display of selected trajectories.D: space-time cube with a single trajectory.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 7: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresExploring and Comparing Individual Trajectories

Projection and clustering of eye trajectories by similarity.

A: a Sammon’s projection of the whole set of trajectories.B: The set is reprojected after removing the outlier.C: a ”table lens” view of trajectory attributes;D: a selected cluster of trajectories in a STC;

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 8: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresInvestigating overall spatial patterns

Space tessellation for generalization and aggregation of eyetrajectories

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 9: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresInvestigating overall spatial patterns

A summary map of eye movements and attention distribution. Thesizes of the circles represent the total time spent in the areas. Thewidth of the violet arrow symbols are proportional to the counts ofeye moves between the areas.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 10: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresExploring Eye Movement Patterns over Time

Summary maps of eye movements by relative time intervals eachrepresenting 10% of the task completion time.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 11: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresExploring Eye Movement Patterns over Time

Summary maps of eye movements by time clusters. The temporalpositions and extents of the clusters are shown on a timeline at thebottom. Using k-mean cluster algorithms with k = 9.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 12: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresExploring Patterns of Users’ Attention

A: The trajectories are represented in a temporal view byhorizontal segmented bars. The colors encode distances toselected AOIs.B: Only trajectory segments satisfying a filter are visible on amap.C: A scatterplot of the counts of the visits of the selectedAOIs against the task completion times.D: The shape of the highlighted trajectory is colose totheoretically optimal.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 13: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresDetermining User’s Difficulties

The map shows the positive and negative differences between thecounts of eye moves (violet and greeen arrow, respectively) andarea visits (red, cyan circles, respectively) for user group 2 and 1.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 14: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresDetermining User’s Difficulties

Repeated visits of the same places for two equivalent tree diagramswith a traditional top-down layout (A,C) and a radial layout (B,D).

A,B: In the temporal view of the eye trajectories, the distancesto previous trajectory points are represented by colorcoding.

C,D: The maps show only the parts of the trajectories wherethe distances to previous points are below 25 pixels.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 15: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Visual Analystics Methods and ProceduresDetermining User’s Difficulties

A: Frequent subsequences discovered by TEIRESIAS areshown in a table view.

B: The frequent subsequences are represented as trajectoriesin a space-time cube.

C: A flow map summarizes the frequent subsequences.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar

Page 16: Visual Analytics Methodology for Eye Movement Studieszwang/schedule/dc11.pdfVisual Analytics Methodology for Eye Movement Studies Gennady Andrienko, Natalia Andrienko, Michael Burch,

Conclusion

Analyzed the eye tracking data with Attention distributions(AOIs) and Attention movements.

Traditional methods for eye tracks analysis versus Geo-VAmethods for movement analysis.

Gennady Andrienko, Natalia Andrienko, Michael Burch, Daniel WeiskopfKSU Graphic & Visualization Lab Seminar