Project Focus
The physical task stress research project is focused on speech analysis and design of speech systems for physical task stress speech.
Data Collection
The physical task stress research project draws from data collected for UT-Scope. The data consist of 72 speakers. Physical task stress was induced with an elliptical/stair-stepper machine. The corpus has 35 sentences of prompted speech for the neutral and physical task stress tasks, and 3 minutes of spontaneous speech during physical task stress.
Above are some examples of audio from the UT-Scope database. The UT-Scope database is not publicly available at this time.
Results
Speech analysis:
- Fundamental frequency changes, formants don't change.
- Open quotient changes
Perception:
- Listeners can identify physical task stress in speech at about 85% accuracy.
Classification:
- A baseline classification system achieves about 12% EER.
Publications
Dissertations/theses:
- Keith W. Godin, "Classification Based Analysis of Speech Under Physical Task Stress", Masters thesis, Univ. of Texas at Dallas, Richardson, TX, USA, Dec. 2009.
Conference papers:
- Keith W. Godin and John H.L. Hansen, "Vowel context and speaker interactions influencing glottal open quotient and formant frequency shifts in physical task stress," in INTERSPEECH-2011, pp. 2945-2948, Florence, Italy, August.
- Keith W. Godin and John H.L. Hansen, "Analysis and Perception of Speech Under Physical Task Stress", in INTERSPEECH-2008, Brisbane, Australia 2008, pp. 1674-1677.
- Sanjay A. Patil and John H.L. Hansen, "Detection of Speech Under Physical Stress: Model Development, Sensor Selection, and Feature Fusion", in INTERSPEECH-2008, pp.817-820.
Personnel
Current:
Former:
- Dr. Sanjay Patil
- Dr. Ayako Ikeno
