Speech analysis for the detection of Parkinson’s disease by combined use of empirical mode decomposition, Mel frequency cepstral coefficients, and the K-nearest neighbor classifier

Abstract

Parkinson’s disease (PD) is one of the neurodegenerative diseases. The neuronal loss caused by this disease leads to symptoms such as lack of initiative, depressive states, psychological disorders, and impairment of cognitive functions as well as voice dysfunctions. This paper aims to propose a system of automatic recognition of Parkinson’s disease by voice analysis. In this system, we are based on a database of 38 recordings, 20 people with Parkinson’s disease and 18 healthy people pronounce the vowel /a/.at first, we have decomposed the vocal signal of each patient by the Empirical Mode Decomposition (EMD), then, we extract from 1 to 12 coefficients of the Mel Frequency Cepstral Coefficients (MFCC), to obtain the voiceprint from each voice sample, we compressed the frames by computing their average value. At the end of the classification, we have used the validation scheme “holdout” as well as the K-nearest neighbor (KNN) classifier, the performance of this classification gives accuracy up to 86,67% when applied to 80% of the database as training data.