Published Paper Details:

Parkinson disease (PD) is a progressive neurodegenerative disease, which has a significant impact on speech production due to motor disabilities. The early and precise diagnosis of PD is a burning clinical issue that indicates the necessity of non-invasive and valid diagnostic methods. This paper proposes a speech-based method of PD detection, and it employs spectral acoustic features and a support vector machine (SVM) classifier. Spectral descriptors such as Mel Frequency Cepstral Coefficients (MFCC) and Mel-spectrogram coefficients are received by acoustic signals in order to capture pathologic variations in the vocal patterns. Experimental findings show that the suggested framework provides an accuracy of 85% exceeding a number of the state-of-the-art approaches. These results suggest that spectral speech biomarkers together with machine learning algorithms provide an effective and non-invasive method of screening PD. Therefore, the suggested method has great potential in aiding the non-invasive diagnosis in the clinical environment.