Development of the Arabic Voice Pathology Database and Its Evaluation by Using Speech Features and Machine Learning...

Area of Science:

• Speech Pathology
• Biomedical Engineering
• Computational Linguistics

Background:

• Automatic voice disorder detection requires specialized databases.
• Ethnic variations in voice characteristics necessitate targeted data collection for accurate diagnosis.
• Existing databases may have limitations that hinder global applicability.

Purpose of the Study:

• To design and develop a comprehensive Arabic Voice Pathology Database (AVPD).
• To address shortcomings of previous voice disorder databases.
• To facilitate research on automatic detection and classification of voice disorders in Arabic speakers.

Main Methods:

• Recorded voice samples including sustained vowels, running speech, and isolated words from Arabic speakers.
• Included perceptual severity ratings for each voice sample.
• Evaluated the database using six speech features and four machine learning algorithms.
• Compared results with the English-language Massachusetts Eye and Ear Infirmary (MEEI) database.

Main Results:

• The AVPD was successfully developed, incorporating unique perceptual severity data.
• The database facilitated the evaluation of voice disorder detection and classification algorithms.
• Performance metrics were obtained for sustained vowels and running speech, enabling comparison with English data.

Conclusions:

• The Arabic Voice Pathology Database (AVPD) is a valuable resource for voice disorder research.
• Developing ethnically specific databases enhances the accuracy and reliability of automatic voice disorder diagnosis.
• The AVPD provides a foundation for improving global voice disorder detection solutions.