An LSTM based DNN Model for Neurological Disease Prediction Using Voice Characteristics

Abstract

INTRODUCTION: A neurological condition known as Parkinson's disease (PD); it affected millions of individuals worldwide. An early diagnosis can help enhance the quality of life for those who are affected with this disease. This paper presents a novel Deep neural network model based on Long Short-Term Memory (LSTM) design for the identification of PD using voice features.
 OBJECTIVES: This research work aims to Identify the presence of PD using voice features of individuals. To achieve this, a Deep neural Network with LSTM is to be designed. Objective of the work is to analyse the voice data and implement the model with good accuracy.
 METHODS: The proposed model is a Deep Neural Network with LSTM.
 RESULTS: The proposed method uses the features gleaned from voice signals for training phase of LSTM model which achieved an accuracy of 89.23%, precision value as 0.898, F1-score of 0.965, and recall value as 0.931and is observed as best when compared to existing models.
 CONCLUSION: Deep Neural Networks are more powerful than ANNs ahd when associated with LSTM , the model outperformed the job of identifying PD using voice data.