Acoustic Signal Based Diagnosis of Neurodegenerative Parkinson's Disease through Machine Learning Approach: A Review

Abstract

The prime objective of this report is to analyze acoustic signals to diagnose Parkinson’s disease. Parkinson’s Disease (PD) is a progressive neurodegenerative disorder that causes damage to the nervous system. PD is the most common neurological condition after Alzheimer’s disease (AD). Language problems, agnosia, apraxia, executive function problems, and emotional control problems are the most common symptoms. In these circumstances, the system performs an acoustic analysis of voice acquisition from PD and Healthy controls through different phonetic and sustained vowel tasks. Preprocessing this acoustic signal using different feature extraction methods such as MFCC (Mel-Frequency Cepstral Coefficients), TQWT (Tunable Q-factor Wavelet Transform), GFCC (Gammatone Frequency Cepstral Coefficients), DWT (Discrete Wavelet Transform), etc. Classifiers like SVM (Support Vector Machines), RF (Random Forest), MLP (Multilayer Perceptron), KNN (K-nearest neighbors) and NN (Neural Network), reveal the existence of acoustic signs compatible with the speech of a PD or AD patient or, on the other hand, the provided audio data proves a healthy pattern. These techniques aim to improve people’s lives by detecting diseases early-stage and identifying patients’ messages to their caregivers. Even though the symptoms of AD or PD have been identified in many people, we require a computer-based assistance in taking care of daily chores and a quality life is more important to neuro disease patients.