A Critical Analysis for Detection of Parkinson’s Disease Based on Voice Signals

Parkinson's disease (PD) is a neurological condition that results in a variety of motor and non-motor symptoms. It is caused due to degeneration of nerve cells in the central nervous system. The motor symptoms include posture instability, gait disorder, bradykinesia and tremors etc. while non-motor symptoms are related to dysphonia, sleep disorder etc. PD is an incurable disease at present. Therefore, an early detection and treatment is required to improve life conditions and is motivation for the research. Various research analyzed motor symptoms to detect PD. However, motor symptoms become visible in late stages. Speech impairment is a symptom which is visible in early stages of PD and therefore remains an open area for research. The state-of-art has utilized various Machine Learning (ML), Deep Learning (DL) and Generative Adversarial Network (GAN) techniques to detect PD in patients. This paper presents a rigorous study of works done between 2018-2023. 14 papers based on detection of PD have been chosen and an analysis of their methodology, limitations and results has been done. Various DL models outperformed the ML models in terms of accuracy but do have certain limitations including imbalanced and noisy datasets and the model’s interpretability. The paper also tries to explore the possibilities of implementation of GAN to improve dataset and proposes a model for PD classification which achieved an accuracy of 94.06. In summary, this study aims to provide an overview of research conducted in the field of speech signal-based PD detection and provides a valuable insight for future research aimed at improving PD detection methods.