Title: Diagnosing Parkinson's disease with speech signal based on convolutional neural network

Authors: Tao Zhang; Yajuan Zhang; Yuyang Cao; Lin Li; Lianwang Hao

Addresses: School of Information Science and Engineering, Hebei Key Laboratory on Information Transmission and Signal Processing, Yanshan University, Qinhuangdao, Hebei, China ' School of Information Science and Engineering, Hebei Key Laboratory on Information Transmission and Signal Processing, Yanshan University, Qinhuangdao, Hebei, China ' School of Information Science and Engineering, Hebei Key Laboratory on Information Transmission and Signal Processing, Yanshan University, Qinhuangdao, Hebei, China ' School of Information Science and Engineering, Hebei Key Laboratory on Information Transmission and Signal Processing, Yanshan University, Qinhuangdao, Hebei, China ' School of Information Science and Engineering, Hebei Key Laboratory on Information Transmission and Signal Processing, Yanshan University, Qinhuangdao, Hebei, China

Abstract: Dysarthria is one of the typical early symptoms of Parkinson's Disease (PD), and that is the basis of diagnosing PD with the speech signal. In this paper, we propose a novel method to analyse the speech signal by Convolutional Neural Network (CNN). At first, the time series signal of speech is converted into spectrograms to represent the time and frequency features in a signal figure; and then, we train the CNN with the spectrograms and their labels from the training set. At last, we test the network precision by the test set of speech signals. The experiments show the accuracy of the method is 91%, which is outperforming the traditional classification for speech signals.

Keywords: diagnosing; Parkinson's disease; speech signal; convolutional neural network; CNN; dysarthria; spectrograms.

DOI: 10.1504/IJCAT.2020.10032598

International Journal of Computer Applications in Technology, 2020 Vol.63 No.4, pp.348 - 353

Received: 14 Dec 2019
Accepted: 25 May 2020

Published online: 19 Oct 2020 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article