Application of artificial intelligence in voice recognition for parkinson’s disease
Main Article Content
Abstract
This study applied artificial intelligence (AI) to recognize the voices of patients with Parkinson’s disease (PD) compared with healthy controls using Vietnamese-language data. A cross-sectional study was conducted at Hanoi Medical University Hospital from 2024 to 2025, involving 20 diagnosed PD patients and 60 healthy controls. Voice samples were recorded, including sustained vowels /a/, /i/, /u/ (≥ 3 seconds, repeated three times), reading a standardized 25-syllable sentence, and 30 seconds of spontaneous speech. Acoustic features analyzed included jitter, shimmer, harmonic-to-noise ratio (HNR), mean and standard deviation of the fundamental frequency (F0), and Mel-frequency cepstral coefficients (MFCC). Two AI models were trained: a Support Vector Machine with a radial basis function kernel (SVM-RBF) and a Convolutional Neural Network (CNN) applied to Mel spectrograms. Performance was evaluated using 5-fold cross-validation. Results: The CNN model achieved an accuracy with an AUC of 0.91, sensitivity of 88%, and specificity of 84%.
Article Details
Keywords
Parkinson’s disease, voice, artificial intelligence, CNN, MFCC, Vietnamese
References
2. Tsanas A, Little MA, McSharry PE, et al. Accurate telemonitoring of Parkinson’s disease progression by noninvasive speech tests. IEEE Trans Biomed Eng. 2012; 59(5): 1264-71. doi:10.1109/TBME.2011.2170335.
3. Orozco‑Arroyave JR, Hönig F, Arias‑Londoño JD, et al. Automatic detection of Parkinson’s disease from continuous speech. J Neurolinguistics. 2016; 37: 141-60. doi:10.1016/j.jneuroling.2016.11.005.
4. Sakar CO, Serbes G, Gunduz A, et al. A comparative analysis of speech signal processing algorithms for Parkinson’s disease classification. IEEE J Biomed Health Inform. 2013; 17(4): 828-34. doi:10.1109/JBHI.2013.2245674.
5. Rusz J, Cmejla R, Ruzickova H et al. Quantitative acoustic measurements for characterization of speech and voice disorders in early PD. J Acoust Soc Am. 2011; 129(1): 350-67. doi:10.1121/1.3514381.
6. Al‑Fatlawi A, et al. Deep CNNs for Parkinson’s disease detection using spectrograms. Comput Methods Programs Biomed. 2021; 199: 105915. doi:10.1016/j.cmpb.2020.105915.
7. Vásquez‑Correa JC, Arias‑Vergara T, Orozco‑Arroyave JR, et al. Convolutional neural networks and transfer learning to classify PD from speech in three languages. Lect Notes Comput Sci. 2019; 11896: 697-708. doi:10.1007/978-3-030-26626-4_68.
8. Shen M, et al. Explainable AI to diagnose early Parkinson’s using voice. Sci Rep. 2025; 15: 5487. doi:10.1038/s41598-025-96575-6.
9. Reddy GP, et al. Artificial intelligence‑based effective detection of Parkinson’s disease using voice. Electronics. 2024; 13(20): 4028. doi:10.3390/electronics13204028.
10. Orozco‑Arroyave JR, Vásquez‑Correa JC, Arias‑Vergara T, et al. Apkinson: The smartphone application to monitor PD patients in the wild. Sensors. 2019; 19(21): 4802. doi:10.3390/s19214802.
11. Rusz J, Krack P, Tripoliti E. Digital speech biomarkers in PD: from prodromal stages to clinical trials. Neurosci Biobehav Rev. 2024; 156: 105922. doi:10.1016/j.neubiorev.2023.105922.
12. Tsanas A, et al. Novel speech signal processing algorithms for PD monitoring. IEEE Trans Biomed Eng. 2014; 61(8): 2180-94. doi:10.1109/TBME.2014.2314574.
13. Ho AK, Iansek R, Marigliani C, Bradshaw JL, Gates S. Speech impairment in PD: a review. Mov Disord. 1998; 13(5): 676-88. doi:10.1002/mds.870130407.
14. Skodda S, Flasskamp A, Schlegel U. Instability of syllable repetition in PD measured by acoustic analysis. J Voice. 2011; 25(6): e265-75. doi:10.1016/j.jvoice.2010.06.009.
15. Dehak N, Kenny P, Dehak R et al. Front‑end factor analysis for speaker verification. IEEE Trans Audio Speech Lang Process. 2011; 19(4): 788-98. doi:10.1109/TASL.2010.2064307.