Related Experiment Videos
The automatic extraction of pitch perturbation using microcomputers: some methodological considerations
J F Deem1, W H Manning, J V Knack
1Memphis State University.
Journal of Speech and Hearing Research
|September 1, 1989
Summary
A new program for automatic jitter extraction (PAEJ) was developed for clinical pitch perturbation measurement. Zero-crossing methods with interpolation yielded the lowest jitter, showing promise for voice analysis.
Area of Science:
- Speech Science
- Acoustic Analysis
- Voice Disorders
Background:
- Accurate measurement of vocal fold vibration perturbations (jitter) is crucial for clinical voice assessment.
- Existing methods for automatic jitter extraction have varying sensitivities and are influenced by multiple factors.
- Development of robust algorithms for fundamental period extraction is needed for reliable jitter analysis.
Purpose of the Study:
- To develop a program for automatic extraction of jitter (PAEJ) for clinical pitch perturbation measurement.
- To compare the relative sensitivity of 12 different algorithms for fundamental period extraction.
- To evaluate the influence of waveform characteristics, noise, and interpolation on jitter measurement accuracy.
Main Methods:
- A microcomputer-based program (PAEJ) was created with 12 algorithms for fundamental period boundary criteria.
- Sine waves were used to compare the sensitivity of extraction procedures regarding waveform peakedness, slope, duration, and noise.
- Pilot data were collected from normal-speaking adults to assess jitter variability.
Main Results:
- Zero-crossing extraction procedures consistently yielded lower jitter values across different sine wave frequencies and sample durations.
- Procedures utilizing positive- or negative-going zero crossings with interpolation demonstrated the lowest jitter measures with sine wave stimuli.
- Pilot data indicated that jitter measures in normal speakers varied based on speaker, vowel, and sample duration.
Conclusions:
- The developed PAEJ program offers multiple algorithms for jitter analysis, with zero-crossing methods showing superior performance.
- Interpolation techniques combined with zero-crossing detection appear most effective for minimizing jitter measurements in controlled stimuli.
- Further research is needed to validate these findings in clinical populations and refine parameters for optimal voice analysis.