Deep Learning-Based Speaker Identification for Individuals with Voice Disorders
Jeyarajasingam S, Charles J, Vigneshwaran P, Hewarathna AI (2025)
Publication Language: English
Publication Type: Conference contribution
Publication year: 2025
Journal
Publisher: Springer
Series: Smart Innovation, Systems and Technologies
City/Town: Singapore
Book Volume: 431
Pages Range: 269-278
Conference Proceedings Title: Information Systems for Intelligent Systems. Proceedings of ISBM 2024, Volume 2
Event location: Bangkok
ISBN: 9789819612093
DOI: 10.1007/978-981-96-1210-9_24
Abstract
Speaker identification is crucial for recognizing individuals based on their voice, a fundamental mode of human communication that enables the expression of thoughts, emotions, and ideas. Voice characteristics can deviate from normal acoustic patterns due to various physiological, pathological, or psychological conditions, including Parkinson’s disease (PD). With the growing use of speech in human–machine interactions and the need for effective audio management in multimedia applications, understanding and identifying these variations is essential. This study investigates speaker identification for individuals with PD through the analysis of audio data. Specifically, it involves extracting 34 distinct features from 400 voice recordings, encompassing spectral, temporal, and chromatic characteristics, which are the predominant acoustic attributes of the audio samples. These features were chosen due to their relevance in reflecting the vocal impairments associated with PD. By employing deep learning models, specifically convolutional neural networks (CNNs) and artificial neural networks (ANNs), the research demonstrates the effectiveness of these methods in identifying speakers with PD. The results show high accuracy rates of 97.50% for CNNs and 96.25% for ANNs, indicating the reliability of these approaches in distinguishing unique audio characteristics associated with PD-affected voices. These findings suggest significant potential for developing automated diagnostic tools and enhancing human–machine interaction systems tailored to individuals with PD. Future work could focus on validating these models with larger, more diverse datasets, comparing them with existing methods, and exploring additional features to further improve identification accuracy.
Involved external institutions
Sri Lanka (LK)How to cite
APA:
Jeyarajasingam, S., Charles, J., Vigneshwaran, P., & Hewarathna, A.I. (2025). Deep Learning-Based Speaker Identification for Individuals with Voice Disorders. In Chakchai So In, Narendra S. Londhe, Nityesh Bhatt, Meelis Kitsing (Eds.), Information Systems for Intelligent Systems. Proceedings of ISBM 2024, Volume 2 (pp. 269-278). Bangkok, TH: Singapore: Springer.
MLA:
Jeyarajasingam, Sharuha, et al. "Deep Learning-Based Speaker Identification for Individuals with Voice Disorders." Proceedings of the 3rd World Conference on Information Systems for Business Management, ISBM 2024, Bangkok Ed. Chakchai So In, Narendra S. Londhe, Nityesh Bhatt, Meelis Kitsing, Singapore: Springer, 2025. 269-278.
BibTeX: Download