Automatic Detection of Parkinson's Disease Using Continuous Wavelet Transform-Based Time-Frequency Domain Analysis of Speech Signals

Parkinson's disease (PD) ranks as the second most prevalent neurological ailment globally, following Alzheimer's disease. The early diagnosis of PD is a significant challenge due to its gradual evolution and the progressive development of its symptoms. Recent research has indicated that alterations in speech can serve as a valuable diagnostic for the timely detection of PD. This paper presents a novel approach for diagnosing PD using speech sounds, utilizing a continuous wavelet transform (CWT). The utilization of CWT was employed to obtain the time-frequency representation (TFR) for each voice recording. From the TFR, entropy and energy are calculated for every frequency component and used as features for the classification of PD-affected speech and healthy individuals' speech. The effectiveness of the proposed framework is verified using the vowels of the GITA database. The support vector machine (SVM) and random forest (RF) are used as a classifier for the classification. Using the vowel /o/, the highest classification accuracy of 88.70% is attained. The results demonstrate that the proposed CWT-based features can accurately diagnose PD using a person's speech.