Vocal Tract Ailments Diagnosis through Voice Features Analysis  

Abstract Category: I.T.
Course / Degree: Natural Language Processing
Institution / University: University of Edinburgh, United Kingdom
Published in: 2008

Thesis Abstract / Summary:

The interpretation of quantitative speech features from the perspective of voice production mechanisms helps extracting information about the pathophysiological aspects of voice disorder. An artificial agent capable of recognizing speech abnormalities is needed for pathological speech assessment. This research work is aimed at finding relationship between voice quality features and vocal tract ailments like Laryngitis and Vocal Nodules. Voice samples from sustained vowel phonatory task have been studied for this purpose previously. Our research is also based on vowels but those are extracted from normal speech samples of paragraph reading. Digital voice samples of persons when sick and healthy were collected. Voice samples were recorded under the supervision of medical practitioners. Their diagnoses were also recorded along with the patient’s voice. Formants analysis was performed to separate out voiced parts of the speech samples for vowels [a],[e] and [o]. The segmented speech samples were listened manually to make sure if segmentation is correct. PRAAT scripts were implemented to generate voice report form different parameters. Integration of the functionalities of different report generation PAART scripts was done by an application developed using Microsoft Visual C++6. We investigated six voice parameters related to pitch, perturbation and loudness for their discriminatory power to diagnose the vocal tract ailments. Effective parameters were found to be fundamental frequency range (F0), jitter, shimmer and mean energy intensity (MEI). Maximum decrease in F0 range was 40Hz and 32Hz for male and female patients of laryngitis for vowels [a]. The maximum increase in jitter for vowel [a] of laryngitis male and female patients was 0.3% and 0.8% and for female vocal nodules patient 1.2%. The maximum increase in shimmer for vowel [a] of laryngitis male and female patients was 4.6% and 5.2% and for female vocal nodules patient 5.4%. The decrease in MEI for vowel [a] of laryngitis male and female patients was 7.3dB and 7.6dB and for female vocal nodules patient 9.1dB. It was noted that MEI values for laryngitis and vocal nodules both male and female patients for vowel [o] followed trend of vowel [a]. The sample segments for vowel [e] were found to be showing fluctuating trend for male and female patients and did not follow a trend to indicate diagnosis.

Thesis Keywords/Search Tags:
NLP, Voice, Speech, A.I., Text, Hearing

This Thesis Abstract may be cited as follows:
No user preference. Please use the standard reference methodology.

Submission Details: Thesis Abstract submitted by Ali Raza Khan from United Kingdom on 02-Jul-2008 17:26.
Abstract has been viewed 3486 times (since 7 Mar 2010).

Ali Raza Khan Contact Details: Email: A.Khan-6@sms.ed.ac.uk

Great care has been taken to ensure that this information is correct, however ThesisAbstracts.com cannot accept responsibility for the contents of this Thesis abstract titled "Vocal Tract Ailments Diagnosis through Voice Features Analysis". This abstract has been submitted by Ali Raza Khan on 02-Jul-2008 17:26. You may report a problem using the contact form.
© Copyright 2003 - 2024 of ThesisAbstracts.com and respective owners.

Copyright © Thesis Abstract | Dissertation Abstracts Thesis Library 2003-2024.
by scope.com.mt @ website design