Synthetic Voice Dataset for Parkinson's Disease Classification using Topological Data Analysis, Random Matrix Theory, and Spectral Features

This release contains the full set of synthetic voice recordings used for training the hybrid TDA–RMT–Spectral framework for Parkinson's disease detection. The archives provided here exceed GitHub's repository file size limits and are therefore distributed as Release assets. The synthetic dataset is designed to approximate the topological, spectral, and random-matrix-driven properties of real pathological and control phonation. These data support the experimental study: Domínguez-Monterroza A., Mateos-Caballero A., Jiménez-Martín A. "Machine Learning for Parkinson's Disease Detection: Analyzing Hybrid Voice Data with Spectral, Topological, and Random Matrix Methods," In review (2025), IEEE Open Journal of Computer Society. Source data Real dataset: The real dataset used as the source for the generative process consists of 150 recordings per group (control and Parkinson), obtained from: C. Gonzalez-Rátiva and E. Nöth, "New Spanish Speech Corpus Database for the Analysis of People Suffering from Parkinson's Disease," Proc. 9th Int. Conf. Lang. Res. Eval., 2014, pp. 342–347. Synthetic dataset: These synthetic recordings were derived from the statistical, spectral, and dynamical characteristics of real phonation data. A total of 1000 synthetic samples per group (control and Parkinson) were generated following the methodology described in: M. Rey-Paredes, C. J. Pérez, and A. Mateos-Caballero, "Time Series Classification of Raw Voice Waveforms for Parkinson's Disease Detection Using Generative Adversarial Network-Driven Data Augmentation," IEEE Open Journal of the Computer Society, vol. 6, pp. 72–83, Jan. 2025, doi: 10.1109/OJCS.2024.3504864. Funding: This paper was supported by the grants PID2021-122209OB-C31, RED2022-134540-T and PID2024-155179NB-C22 funded by MICIU/AEI/10.13039/501100011033.