Related Experiment Videos
A study of speech fractal dimensions
Acta Oto-Laryngologica
|May 13, 1999
Summary
The fractal dimension of vowel sounds was analyzed using the box counting method. While not fractal, vowel signals may offer unique signatures for each sound, potentially aiding clinical speech analysis.
Area of Science:
- Acoustics
- Signal Processing
- Phonetics
Background:
- Fractal dimension (D) quantifies signal complexity and fluctuations.
- Understanding speech signal characteristics is crucial for various applications.
Purpose of the Study:
- To investigate the fractal dimension of French vowels.
- To explore the potential of signal roughness for distinguishing vowels and clinical applications.
Main Methods:
- Utilized 16 kHz time sampling and the box counting method.
- Analyzed stationary French vowel phonemes ([a], [e], [i], [o], [y]).
- Data collected from 10 males and 10 females, each pronouncing vowels four times.
Main Results:
- The stationary vowel signal was found not to be fractal.
- Identified potential for unique signal signatures at the smallest scale for each vowel.
- Categorization based on these signatures was not entirely significant.
Conclusions:
- The box counting method quantifies signal roughness, offering potential clinical applications.
- Vowel signatures could be integrated into speech signal processing for cochlear implants, especially for individuals with dysphonia.