Estimating dysphonia severity in continuous speech: application of a multi-parameter spectral/cepstral model
Shaheen N Awan1, Nelson Roy, Christopher Dromey
1Department of Audiology & Speech Pathology, Bloomsburg University of Pennsylvania, Bloomsburg, PA 17815-1301, USA. sawan@bloomu.edu
Clinical Linguistics & Phonetics
|November 7, 2009
Summary
This study identified key acoustic measures to predict voice disorder (dysphonia) severity in continuous speech. A model using cepstral peak prominence and spectral energy ratios accurately estimated severity, aiding objective treatment assessment.
Area of Science:
- Speech-Language Pathology
- Acoustic Analysis
- Voice Disorders
Background:
- Muscle tension dysphonia (MTD) is a common voice disorder.
- Objective measures are needed to assess MTD severity and treatment outcomes.
- Auditory-perceptual analysis is subjective and time-consuming.
Purpose of the Study:
- To identify spectral/cepstral acoustic methods for predicting dysphonia severity in continuous speech.
- To develop an objective measure for treatment outcomes in MTD.
- To correlate acoustic measures with auditory-perceptual ratings of dysphonia.
Main Methods:
- Analysis of 104 pre- and post-treatment continuous speech samples from females with MTD.
- Auditory-perceptual rating using a 100-point visual analogue scale (VAS).
- Acoustic analysis using spectral/cepstral measures, including cepstral peak prominence (CPP) and low-to-high frequency spectral energy ratios.
Main Results:
- A three-factor regression model (CPP, mean and std dev of low/high frequency spectral energy ratio) strongly correlated with perceived dysphonia severity (R = .85, R2 = .73).
- Predicted severity ratings closely matched perceptual ratings (mean difference < 6 VAS points).
- The model demonstrated accuracy across diverse MTD voice types and severities.
Conclusions:
- A multi-parameter acoustic model using spectral/cepstral measures shows promise as an objective dysphonia severity measure.
- This model can objectively assess treatment outcomes in continuous speech for MTD.
- Acoustic analysis offers a reliable alternative to subjective perceptual ratings for voice disorders.