Speech Signal Analysis Using Gammatone-Frequency Cepstral Coefficient For Parkinson's Disease Prediction
Pandit Vivek Kumar Pandey, Sitanshu Sekhar Sahu · Apple Academic Press eBooks · 2025
Parkinson’s disease (PD) is a neurodegenerative condition characterized by a lack of dopamine hormone secretion as a result of nerve cell damage in the human brain. The symptoms of PD may be alleviated with the aid of early identification and therapy. Patients with PD have a condition called hypokinetic dysarthria, which affects every element of their speech, including their voice production, breathing, phonation, articulation, nasality, and prosody. Recent research on PD has extracted elements from speech abnormalities as a precursor for PD identification. This is because people with PD have changes and impairments in their speech characteristics at an early stage of the disease. We employed a PC_GITA dataset in order to construct an evaluation of speech problems that may be used to identify patients with PD. The mel-frequency cepstral coefficient (MFCC) is the most commonly used feature for evaluating PD speech. In this chapter, the gammatone-frequency cepstral coefficients (GFCC) feature is explored to identify Parkinson’s speech. Random forest (RF) and 132support vector machine (SVM) with multiple kernel types were utilized for classification. The SVM classifier predicts with better accuracy than the RF classifier. In Vowel, the vowels /i/ and /e/ performed with greater than 80% accuracy. Vowel /i/ has gotten 86.67% accuracy in SVM; in Words, the words /petaka/, /braso/, and /name/ performed with greater accuracy between 60 and 80%. Word /braso/ has gotten 75% accuracy and in Sentence, between 60 and 80% accuracy performed in the GFCC. Sentence/laura/ and /loslilibros/ performed with an accuracy 75%. Overall, GFCC demonstrated a 0–20% improvement in accuracy over MFCC.