Using multimodal speech production data to evaluate articulatory - - PowerPoint PPT Presentation

using multimodal speech production data to evaluate
SMART_READER_LITE
LIVE PREVIEW

Using multimodal speech production data to evaluate articulatory - - PowerPoint PPT Presentation

Data Registration Animation Evaluation Using multimodal speech production data to evaluate articulatory animation for audiovisual speech synthesis Ingmar Steiner Korin Richmond Slim Ouni N I V E U R S E I H T T Y O H F G R


slide-1
SLIDE 1

Data Registration Animation Evaluation

Using multimodal speech production data to evaluate articulatory animation for audiovisual speech synthesis

Ingmar Steiner Korin Richmond Slim Ouni

T H E U N I V E R S I T Y O F E D I N B U R G H

University College Dublin CSTR LORIA & Trinity College Dublin University of Edinburgh Universit´ e de Lorraine

Vienna, September 21, 2012

slide-2
SLIDE 2

Data Registration Animation Evaluation

Motivation

Data-driven animation for speech articulators

slide-3
SLIDE 3

Data Registration Animation Evaluation

Motivation

Data-driven animation for speech articulators within the vocal tract

slide-4
SLIDE 4

Data Registration Animation Evaluation

The mngu0 Corpus

Multimodal speech corpus

  • one male speaker of British English
  • electromagnetic articulography (EMA)
  • magnetic resonance imaging (MRI)
  • dental cast scans

http://mngu0.org/

slide-5
SLIDE 5

Data Registration Animation Evaluation

Electromagnetic articulography

slide-6
SLIDE 6

Data Registration Animation Evaluation

MRI

volumetric vocal tract imaging

slide-7
SLIDE 7

Data Registration Animation Evaluation

MRI

manual regions of interest (ROIs)

slide-8
SLIDE 8

Data Registration Animation Evaluation

MRI

isosurfaces within ROIs

slide-9
SLIDE 9

Data Registration Animation Evaluation

Dental scans

vertex count 927 282 (maxilla), 836 892 (mandible)

slide-10
SLIDE 10

Data Registration Animation Evaluation

Dental scans

deduplication: vertex count 154 549 (maxilla), 139 484 (mandible)

slide-11
SLIDE 11

Data Registration Animation Evaluation

Dental scans

decimate (5 %)

slide-12
SLIDE 12

Data Registration Animation Evaluation

Dental scans

vertex count 7729 (maxilla), 6976 (mandible)

slide-13
SLIDE 13

Data Registration Animation Evaluation

Palate contour

slide-14
SLIDE 14

Data Registration Animation Evaluation

Palate contour

slide-15
SLIDE 15

Data Registration Animation Evaluation

Model rigging

EMA motion capture data

slide-16
SLIDE 16

Data Registration Animation Evaluation

Model rigging

maxilla/mandible track ref/jaw coils

slide-17
SLIDE 17

Data Registration Animation Evaluation

Tongue mesh retopology

crude isosurface

slide-18
SLIDE 18

Data Registration Animation Evaluation

Tongue mesh retopology

tesselation from MRI voxels

slide-19
SLIDE 19

Data Registration Animation Evaluation

Tongue mesh retopology

simple cage

slide-20
SLIDE 20

Data Registration Animation Evaluation

Tongue mesh retopology

“shrinkwrapped” to isosurface

slide-21
SLIDE 21

Data Registration Animation Evaluation

Tongue mesh retopology

Catmull-Clark subdivision

slide-22
SLIDE 22

Data Registration Animation Evaluation

Tongue mesh retopology

smooth, tongue-shaped mesh with simple topology

slide-23
SLIDE 23

Data Registration Animation Evaluation

Tongue rigging

static mesh

slide-24
SLIDE 24

Data Registration Animation Evaluation

Tongue rigging

spline (NURBS path)

slide-25
SLIDE 25

Data Registration Animation Evaluation

Tongue rigging

modified by hooks tracking tongue coils

slide-26
SLIDE 26

Data Registration Animation Evaluation

Tongue rigging

armature follows spline through inverse kinematics (IK)

slide-27
SLIDE 27

Data Registration Animation Evaluation

Tongue rigging

tongue mesh deformed by armature

slide-28
SLIDE 28

Data Registration Animation Evaluation

Animation

slide-29
SLIDE 29

Data Registration Animation Evaluation

Animation

slide-30
SLIDE 30

Data Registration Animation Evaluation

Animation

slide-31
SLIDE 31

Data Registration Animation Evaluation

Animation

slide-32
SLIDE 32

Data Registration Animation Evaluation

Vertex tracking

T1 T2 T3

  • 2

2 4 6 20 30 40 50 60

  • 10
  • 5

5 x y z 1 2 3 1 2 3 1 2 3

time (s) position (mm)

trajectory EMA vertex

slide-33
SLIDE 33

Data Registration Animation Evaluation

Vertex tracking

T1 T2 T3

  • 2

2 4 6 20 30 40 50 60

  • 10
  • 5

5

r=0.98 r=0.99 r=0.99 r=0.92 r=0.98 r=0.92 r=0.89 r=0.96 r=0.91

x y z 1 2 3 1 2 3 1 2 3

time (s) position (mm)

trajectory EMA vertex

slide-34
SLIDE 34

Data Registration Animation Evaluation

Conclusion

Skeletal animation of articulatory movements from speech production data seems promising, but depends on

  • model topology
  • data quality
  • registration (incl. posture effect)