Quiz

1
Quiz

• 1. What sound localization technique do people
  use to segregate between auditory information?
• (a) Cuboid Space Differentiation
• (b) Auditory Super-Normal
• (c) Equidistant Localization
• (d) Cocktail Party Effect
• 2. Head Related Transfer Function or HRTF encodes
  a sound signal to create sound spatialization.
  HRTF encodes the signal by simulating the parts
  of ones body that filter a sound signal. What
  parts of the body do this task?
• (a) head, torso, pinnae (i.e. outer part of
  ears)
• (b) inner part of ears, nose, throat
• (c) head, shoulders, knees, toes
• (d) inner part of ears, nose, head

2
Quiz

• 3. What size relation was the visual space in
  relation to the auditory space?
• (a) Both are the same size.
• (b) The visual space is larger than the auditory
  space.
• (c) The auditory space is larger than the visual
  space.
• (d) This fact was not mentioned in the paper.
• 4. Which of the stereo spatial stimuli was least
  correctly recognized?
• (a) Depth stimuli
• (b) Elevation (i.e. vertical deviation) stimuli
• (c) Azimuth (i.e. horizontal deviation) stimuli
• (d) Another stimuli not mentioned in (a), (b),
  or (c)

3
Mapping an Auditory Space onto a Graphical User
Interface

• Michael J. Evans

Presented by Allan Spale EECS 578
4
Information Presentation

• Information presentation typically uses a visual
  interface
• Integration of audio to convey information on the
  desktop computer
• GUI-based file system actions
• Speech audio
• Teleconferencing and videoconferencing
• Computer games

5
Potential Benefits of Spatial Audio Reproduction

• Sound localization plays a role in choosing which
  audio to listen to
• Cocktail Party Effect
• Spatial hearing in noisy environments
• Applications that could use spatialized sound
• Operating system GUIs
• Teleconferencing
• Interfaces for the blind

6
System Infrastructure

• Model audio signal using Head-Related Transfer
  Function (HRTF)
• Simulates how a person filters audio using the
  head, torso, and pinnae
• Use frontally-placed loudspeakers
• Should cancel left-right crosstalk
• Keep the user sitting still in an ideal area
  for obtaining sound spatialization

7
Problems with the Proposed System Infrastructure

• Small head movements may disrupt perception of
  spatial audio
• Limited to use by one listener
• HRTF measurements
• Time-consuming
• Requires anechoic chamber
• Accurate calculations of azimuth and elevation
• HRTFs vary according to an individual

8
Visual and Auditory Space

• Visual Implementation
• 15-inch monitor
• Visible display 0.28 x 0.21 meters
• Visual space (approx.) 12? above and below
  horizontal plane and 16? on either side of the
  median plane

9
Visual and Auditory Space

• Visual Implementation
• Virtual cuboid space for subjective depth
  perception
• Resizing items similar to bringing items nearer
  to or farther away from the user
• Active items will be nearest to the user
• GUI items can be moved within the cuboid

10
A Cuboid Space Mapped onto a Virtual Display
11
Visual and Auditory Space

• Audio Implementation
• Continuous transaural algorithm
• JBL loudspeakers placed 30? either side of the
  center of the monitor
• Auditory space mapped on visual cuboid
• Sound pressure level has an inverse-square
  relationship with increasing distance

12
Exaggerated Spatialization

• People have a greater auditory acuity than visual
  acuity
• Map display onto an exaggerated auditory space
• Benefits of exaggerated spatialization
• Permit users to experience a greater range of the
  intended spatial effect

13
Diagram of anExaggerated Spatialization
14
Problems with Visual and Auditory Conflicts

• Large differences in spatial mismatch has user
  perceive separate events
• People are able to tolerate some directional
  disparity
• Conflicts of audio and visual items usually are
  resolved with considering the location of the
  visual stimulus

15
Pilot Test Design

• Goal
• Perceptual tests will attempt to confirm that
  visual and auditory interfaces remain congruent
  to users, even when the auditory interface space
  is exaggerated.

16
Pilot Test Design

• Testing subject
• Sat in chair 0.5 meter away directly facing the
  screen
• Adjustable floor stands hold speakers
• Tilted at 30? in relation to the user heads
  median plane
• Visual area
• 0.28 x 0.21 meters 16? azimuth, 12? elevation
• Auditory area (2x visual area)
• 0.56 x 0.42 meters 32? azimuth, 24? elevation

17
Schematics of the Pilot Perceptual Tests
18
Pilot Test Design

• Visual GUI
• Black background
• Rectangles
• Blue and red
• Eighteen possible positions in visual space
• Nine possible positions each on the front surface
  and the rear surface
• Depth mapped using size

19
Pilot Test Design

• Visual Stimuli
• Twelve in total with each containing a red
  rectangle and a blue rectangle
• Six specifued layouts
• Chosen to present extremes of variation in
  each dimension and the combination of
  dimensions.

20
Layouts and Definitions of the Visual Artifacts
21
Pilot Test Design

• Auditory Stimuli
• Non-contextual pair of sentences spoken by
  one of four talkers...
• Six sentence pairs associated with A through F
  produced non-spatialized sounds
• Six sentence pairs associated with A through F
  for produced spatialized sounds
• Rectangle usage
• Subject had to indicate where the sound was
  originating, which rectangle or no rectangle at
  all

22
Testing Information

• Users
• Untrained, one woman and five men
• No hearing problems, normal color vision
• Tested on the twelve audiovisual stimuli
• Not informed that tests involved spatialized
  audio
• Testing Location
• 3 x 2 x 3 meter office, furnished and carpeted

23
Table of Test Results
Check mark Correct rectangle Dash Neither
rectangle X Wrong rectangle
24
Discussion of Results

• Support exaggerating the auditory interface
• Mono references varied greatly among subjects
• Stimulus B
• Reduced success rate compared to others
• Elevation-only stimulus related to individual
  nature of HRTFs
• Stimuli in the median plane are unable to use
  localization cues from the comparison of left and
  right ear signals

25
Discussion of Results

• Stimulus C
• When differing in elevation and depth, subjects
  can recognize sound location
• Stimuli D, E, and F
• 100 correct recognition
• Stimuli D, E, and F
• Subjects usually chose the incorrect rectangle

26
Conclusions

• Spatial audio integrated in conventional GUI
• Map virtual cuboid of visual space onto an
  exaggerated audio space
• Use transaural cancellation with monitor-side
  loudspeakers
• Congruence maintained between visual and auditory
  artifacts

27
Conclusions

• Problems
• Auditory elevation cues are very individualized
• More than one artifact in the median plane causes
  difficulty in sound localization
• Future work
• Determine limitations of exaggerated auditory
  interfaces
• Use less generalized interfaces
• Range of different exaggerated auditory interfaces

28
Some ReferencesListed in the Paper

• Integrating non-speech audio into interfaces
• Reference 4,5
• Sound localization in teleconferencing
  applications
• Reference 12
• Transaural technique
• Reference 19

29
Questions and Comments