A NOVEL VOICE-BASED SYSTEM FOR PARKINSON’S DISEASE DETECTION USING RNN-LSTM

Abstract

PD (Parkinson's Disease) is a neurological illness that develops with time and causes motor and non-motor symptoms. A lot of PD patients have trouble moving normally in the early phases of the illness. Vocal disorders are among the most prevalent symptoms. Latest PD detection investigations have concentrated on diagnostic methods on the basis of vocal problems, which are an excitingly new area of study with a lot of potential. For a range of prediction problems that are troubling medical practitioners, Deep Learning (DL) has gained popularity recently. In this study, RNN-LSTM is combined with numerous architectures to develop more accurate prediction models for the detection of PD on the basis of feature analysis of various patient speech samples. Importantly, Deep Neural Networks have become the best classification tool for PD detection even without the use of a feature selection strategy. RNN-LSTM was then fine-tuned, resulting in an accuracy of 98.772 percent.

Keywords

• classification
• deep neural networks
• parkinson’s disease
• deep learning