Parkinson's disease detection from voice using artificial intelligence techniques: a review
Date
2024
Authors
Ali, M.H.
Mohammed, S.L.
Al Naji, A.
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Conference paper
Citation
AIP Conference Proceedings, 2024, vol.3232, iss.040010, pp.1-7
Statement of Responsibility
Conference Name
5th Scientific Conference for Electrical Engineering Techniques Research, EETR 2024 (15 Jun 2024 - 16 Jun 2024 : Baghdad)
Abstract
Parkinson's disease (PD) is a long-term neurodegenerative disorder affecting the human motor system, resulting in many motor and non-motor characteristics. The disease arises as a result of the disappearance of neurons that generate dopamine and are located in the substantia nigra region of the central nervous system (CNS), which affects the human body. This reduction in neuronal count contributes to diminished dopamine levels, consequently giving rise to both motor and non-motor symptoms. Motor dysfunctions observed in individuals with PD encompass voice, writing, and ambulation challenges. Non-pharmacological interventions that incorporate the use of voice analysis possess the capability to enhance the overall well-being of patients through the mitigation of speech impairments. Consequently, models based on speech analysis are progressively utilized for the remote monitoring and diagnosis of Parkinson's disease. Nonetheless, accurately interpreting speech signals remains a formidable obstacle in PD classification. Detecting Parkinson's disease from voice involves a multifaceted process. Researchers typically collect audio recordings from individuals both with and without Parkinson's disease, then extract a variety of acoustic features such as pitch, jitter, shimmer, and formant frequencies. These features serve as inputs for machine learning models, which are trained to classify recordings as either indicative of Parkinson's or not. Performance evaluation metrics, including accuracy and sensitivity, validate model effectiveness before deployment for diagnostic purposes. Various artificial intelligence approaches have been used successfully in academics for classification purposes, presenting the opportunity for timely diagnosis. The main purpose of this study is to extensively investigate machine learning and deep learning methodologies used to classify Parkinson's disease by analyzing speech data. This review study suggests that artificial intelligence approaches, specifically deep learning, hold significant promise to assist neurologists and data scientists in the decision-making procedures linked to detecting Parkinson's disease
School/Discipline
Dissertation Note
Provenance
Description
Access Status
Rights
Copyright 2024 The author(s).