Performance Assessment Methodology: Task Dependent Evaluation of Display Systems

Abstract

As the focus of this research, a new

methodology -- human Performance

Assessment Methodology (PAM), is

introduced. PAM provides a quantitative

basis for evaluating display image quality

based on the visual events that occur in a

task. The PAM approach identifies the

visual events, decisions, and actions for a

display system. To support PAM, a

theoretical model, the Model of Visual

Events (MOVE), is proposed for describing

the relationship between visual events,

decisions, and actions. MOVE describes

four categories of perceptual decisions (i.e.,

detect, identify, discriminate, and evaluate)

associated with visual events. Formal

efficiency metrics are introduced in PAM to

describe performance at the visual event,

task, and network levels. Using PAM, an

efficiency model was created for one visual

display parameter (i.e., luminance), one

decision type (i.e., detection) and one

dependent variable (i.e., visual angle). Two

experiments were accomplished to examine

the validity of PAM. A two-factor mixed

design was employed for both experiments,

where decision type was varied

between-subjects and visual display

parameter (i.e., luminance or sharpness)

was varied within-subjects. In the first

experiment, luminance was varied across

four levels (3.2, 4.5, 8.6, 16.5 cd/m2) for

two decision types (detection and

identification). In the second experiment,

three levels of sharpness (50% spot width -

0.508, 0.711, 0.864 mm) were combined

factorially with two decision types

(detection and identification). In both

experiments, participants visually 'walked

down a path' and either detected or

identified visual targets presented on the

screen. Time-to-target and subjective

responses were measured for each study.

The results of the first experiment show that

time-to-target and subjective rating

significantly change as a function of

luminance. For the sharpness variable in the

second experiment, a significant difference

was found for time-to-target while

subjective rating was non-significant. In both

studies, participants detected visual targets

quickly, but required more time to identify

targets. Using the PAM, functional

relationships for luminance and sharpness

were determined for detection and

identification decisions. When detection

data from the current study were contrasted

with previous detection data, general

agreement was found between the data sets.

This research defines PAM and shows its

utility for modeling the functional

relationships among visual parameters.

Further research is needed to validate and

refine the PAM approach.