Coarticulation Analysis of Dysarthric Speech

1
Coarticulation Analysis of Dysarthric Speech

• Xiaochuan Niu,
• advised by Jan van Santen

2
Outline

• Goal of the Dysarthria Project
• Problems
• Hypotheses
• Analysis approach
• Results
• Conclusions

3
Goal of the Dysarthria Project

• Dysarthria Motor speech impairment
• Dysarthric speech
• Normal speech
• Improve the intelligibility of Dysarthric speech

4
Problems

• Previous results of intelligibility test (by
  Hosom, Kain, et al.)
• Improvement potential
• Dysarthric 68 --gt Normal 99
• Spectral feature replacement 87
• Baseline transformation system
• GMM Linear transformation
• NO improvement 67
• Poor spectral separability of dysarthric speech

5
Hypotheses

• Poor spectral separability caused by
• H1 - Target shift
• Dysarthric speakers develop special vocal-tract
  configurations for certain phonemes.
• H2 Coarticulation effect
• Degree of context influence on articulation is
  greater in dysarthric speech
• H3 - Random variation
• Dysarthric speakers can not repeat the target
  vocal-tract configuration accurately

6
Analysis approach

• Acoustic measure
• Speech data
• Coarticulation model
• Estimation method

7
Acoustic measure

• Formants
• Natural frequencies of certain vocal-tract
  configurations
• Assumption each phoneme has a target formant
  pattern
• Formant trajectories
• Dynamic characteristics of articulation

8
Acoustic measure (example)

• Formant trajectories of CVC segments

9
Speech data

• Utterance
• One dysarthric speaker and one normal speaker
  from a dysarthric database
• 74 nonsense sentences per speaker
• Phoneme
• Manually labeled with time alignments
• Formants
• First three formant frequencies at the midpoints
  of vowels in CVC segments
• Automatically extracted and manually checked

10
Coarticulation model

• Notations
• Observed formant vectors
• Target formant vectors
• Coarticulatory factors

11
Estimation method

• Given N samples of observed formant vectors
  at midpoints of vowels in CVC segments, assume
  target formant-vectors are known,

• With and fixed, jointly estimate
  target formant-vectors from equations

12
Results H1

• Vowel space (/i, _at_, A, u/) dysarthric normal

13
Results H2

• Coarticulation effects

14
Conclusions

• An approach to decompose the contributions of
  three factors
• target shift / coarticulatory effect / random
  variation
• Practical aspects of the approach
• Initial targets
• Target constraints in estimation
• Future work
• Analysis of the entire trajectory
• Apply results in the transformation system