Datum / čas
Date(s) - 29.05.
10:30
Kategorie
Lecture of Prof. Abdulnasir Hossen
UNESCO Chair on Artificial Intelligence,
Sultan Qaboos University,
Oman
WHERE: CIIRC ČVUT, room 101 – ground floor
WHEN: 29. 5. 2026
Classification of the Severity of Speech Disorder in Patients with Parkinson Disease using Artificial Intelligence Techniques
Anstract: This presentation investigates the use of Artificial Intelligence (AI), Machine Learning (ML), and Digital Signal Processing (DSP) techniques for the classification of speech disorder severity in patients with Parkinson’s Disease (PD). Parkinson’s Disease is one of the most common neurological disorders worldwide and is often associated with hypokinetic dysarthria, a speech impairment affecting nearly 90% of PD patients. Traditional assessment methods for dysarthria severity rely heavily on auditory evaluation by specialists, which can be time-consuming, expensive, and subject to variability and misdiagnosis.
The presentation proposes automated AI-based systems to classify dysarthria severity into both five-class and three-class categories using speech signal analysis. Speech recordings collected from the Hospital of the University of Kiel, Germany, were analyzed using several feature extraction techniques, including Linear Predictive Coding (LPC), Mel-Frequency Cepstral Coefficients (MFCCs), and Discrete Packet Wavelet Decomposition (DPWD). Multiple machine learning models such as Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Neural Networks (NN) were implemented and evaluated.
The work also investigates the influence of gender and speech type on classification accuracy, including sustained vowels, repeated syllables, spontaneous speech, and short expressions. Various voting approaches, including Subject-Method Voting, Subject-Model Voting, Segment-Subject Voting, and Segment-Subject-Model Voting, were applied to improve classification performance and robustness.
Experimental results demonstrate that the proposed AI-based framework achieved high classification accuracies, particularly when combining DPWD features with advanced voting strategies. The findings indicate that automated AI systems can significantly improve the assessment of speech disorders associated with Parkinson’s Disease while supporting early diagnosis, monitoring disease progression, and evaluating speech therapy effectiveness.
The presentation concludes that AI-driven speech analysis has strong potential as a non-invasive, low-cost, and efficient clinical decision-support tool for neurological disease assessment and personalized healthcare applications.