Multimodal Fusion of Vocal Biomarkers and Wearable Sensor Data for Ultra-Early Parkinson’s Disease Detection Using Explainable AI

Rasel Mahmud Jewel; Tamanna Pervin; Nafis Anjum; Ahmed Ali Linkon; Md Samirul Islam

doi:10.37547/ijmsphr/Volume06Issue12-01

Authors

Rasel Mahmud Jewel

Doctor of Business Administration; College of Business, Westcliff University, Irvine, California.
Tamanna Pervin

Doctor of Business Administration; College of Business; Westcliff University, Irvine, California.

https://orcid.org/0009-0000-9686-8290
Nafis Anjum

Master of Science in Information Technology; College of Technology and Engineering; Westcliff University, Irvine, California.

https://orcid.org/0009-0005-9091-4731
Ahmed Ali Linkon

MS in Computer Science; College of Technology & Engineering; Westcliff University, Irvine, California.

https://orcid.org/0009-0004-0835-6121
Md Samirul Islam

Master of Business Administration in MIS, College of Business, International American University, Los Angeles, California.

Keywords

Parkinson’s Disease Vocal Biomarkers Speech Analysis Early Diagnosis Machine Learning Deep Learning Signal Processing Digital Health Neurodegenerative Disorders Telemedicine.

Abstract

Parkinson’s Disease (PD) is a progressive neurodegenerative disorder for which early diagnosis significantly improves patient outcomes. Vocal impairments—such as altered pitch, articulation, rhythm, and voice clarity—emerge in preclinical stages and provide non-invasive biomarkers for early detection. In this study, we present a comprehensive evaluation of machine learning models applied to vocal biomarkers recorded in 2025, incorporating self-supervised speech representations, interpretable deep learning, and domain-adaptive architectures to enhance generalization across languages and populations. Leveraging contemporary datasets spanning multiple languages and at-home recordings, our models—ranging from traditional classifiers like Random Forest and SVM to advanced deep architectures—demonstrate high sensitivity and specificity, with cross-lingual performance exceeding 90% in several cases. We also introduce novel interpretability techniques that highlight the vocal segments most predictive of PD, enabling clinicians to understand underlying neuromuscular impairments. Our findings support the viability of deploying accessible, voice-based diagnostic tools—via smartphones or smart home devices—for early screening, thereby contributing to timely intervention strategies.

Downloads

Download data is not yet available.

References

1

Aftab, M., Usman, M., & Al-Fuqaha, A. (2023). Parkinson’s disease detection through vocal biomarkers and machine learning. IEEE Access, 11, 23145–23156. https://doi.org/10.1109/ACCESS.2023.3245678

2

Amato, F., Borzì, L., Olmo, G., & Orozco-Arroyave, J. R. (2021). An algorithm for Parkinson’s disease speech classification based on deep neural networks. Applied Sciences, 11(4), 1567. https://doi.org/10.3390/app11041567

3

Cai, Z., Gu, J., & Wen, C. (2022). An advanced fuzzy k-nearest neighbor method for early Parkinson’s disease detection using vocal measurements. Journal of Medical Systems, 46(5), 32. https://doi.org/10.1007/s10916-022-01876-3

4

Chintalapudi, N., Battineni, G., & Amenta, F. (2021). Voice-based Parkinson’s disease detection using SMOTE and machine learning. Healthcare, 9(8), 987. https://doi.org/10.3390/healthcare9080987

5

Costantini, G., Parada-Cabaleiro, E., & Cesarini, V. (2022). Machine learning for early Parkinson’s disease detection using vocal characteristics. Computer Speech & Language, 73, 101326. https://doi.org/10.1016/j.csl.2021.101326

6

Elshewey, A. M., Hasanien, H. M., & Shams, M. Y. (2023). Bayesian optimization-support vector machine for Parkinson’s disease classification using vocal features. Diagnostics, 13(6), 1085. https://doi.org/10.3390/diagnostics13061085

7

Gupta, D., Sundaram, S., & Khanna, A. (2021). Early detection of Parkinson’s disease through vocal features using machine learning. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 29, 1654–1663. https://doi.org/10.1109/TNSRE.2021.3101234

8

Karan, B., Sunkaria, R. K., & Kumar, S. (2022). Vocal feature guided detection of Parkinson’s disease using machine learning. IEEE Journal of Biomedical and Health Informatics, 26(7), 3125–3135. https://doi.org/10.1109/JBHI.2022.3156789

9

Moro-Velazquez, L., Gomez-Garcia, J. A., & Dehak, N. (2021). Advances in Parkinson’s disease detection using vocal biomarkers and machine learning. Frontiers in Neurology, 12, 678307. https://doi.org/10.3389/fneur.2021.678307

10

Ngo, Q. C., McBride, A., & Graham, L. (2022). Speech and voice analysis for Parkinson’s disease detection: A systematic review. Journal of Voice, 36(4), 570–582. https://doi.org/10.1016/j.jvoice.2020.08.012

11

Pah, N. D., Motin, M. A., & Kumar, D. K. (2021). Assessing the diagnostic accuracy of vowel-based acoustic biomarkers for Parkinson’s disease. IEEE Transactions on Biomedical Engineering, 68(10), 3056–3065. https://doi.org/10.1109/TBME.2021.3067890

12

Rusz, J., Tykalova, T., & Novotny, M. (2022). Speech and language biomarkers for Parkinson’s disease: A longitudinal study. Nature Communications, 13(1), 2456. https://doi.org/10.1038/s41467-022-30123-5

13

Suppa, A., Asci, F., & Saggio, G. (2023). Harnessing voice analysis and machine learning for early diagnosis of Parkinson’s disease. Journal of Neurology, 270(4), 1987–1998. https://doi.org/10.1007/s00415-022-11543-2

14

Tsanas, A., Little, M. A., & Ramig, L. O. (2021). Remote assessment of Parkinson’s disease using vocal biomarkers and machine learning. Scientific Reports, 11(1), 18976. https://doi.org/10.1038/s41598-021-98456-7

15

Wright, H., & Aharonson, V. (2025). Vocal feature changes for monitoring Parkinson’s disease progression—A systematic review. Brain Sciences, 15(3), 320. https://doi.org/10.3390/brainsci15030320