<![CDATA[The manual diagnosis of dysarthria is often time-consuming and requires the expertise of trained specialists, which can delay early intervention and treatment. This study aims to develop an automated detection system to improve diagnostic accuracy and efficiency. Mel-Frequency Cepstral Coefficients (MFCC) are used as the primary features, and three classification models are evaluated: Support Vector Machine (SVM), Multilayer Perceptron (MLP), and a stacking ensemble that combines both. The evaluation is conducted on a dataset of 240 audio samples. Experimental results show that the stacking ensemble achieves the highest performance, with an accuracy of 97.92%, surpassing SVM (95.83%) and MLP (93.75%). These findings highlight the significant potential of voice-based classification to accelerate dysarthria diagnosis, thus supporting clinical screening and speech therapy applications.]]>
Copyrights © 2025
