Rahul Soangra, Michael Shiraishi
Introduction: Parkinson's disease is a disorder in the central nervous system that causes tremors, abnormal gait and balance, and muscle rigidity due to loss of function in parts of the brain. Traditionally, Parkinson's is identified by the physical symptoms seen in a patient's gait and motor skills, but irregular speech patterns (hypokinetic dysarthria) is one of the first symptoms to be derived from the disease. Objective: The goal of this ongoing study is to use Mel Frequency Centrum Coefficients (MFCC), to diagnose Parkinson's in the early stages by identifying hypokinetic dysarthria. MFCC's process speech recognition patterns and produce a frequency scale that software can use to identify differences between subjects. This was first performed using healthy male and female subjects, in which MFCC's successfully identified the gender of the individual speaking. Using this finding, we can apply the process to identify Parkinson's disease in early stages. Symptom identification using machine learning can increase the patient's lifespan through early detection and early therapies. Methods: The patients were asked to perform various vocal tests, as well as normal speaking in order to obtain the vocal range frequencies of each test to isolate MFCC's to identify Parkinson's via biomarkers in voice patterns. Conclusion: Parkinson's is most commonly identified when the patient develops abnormal gait patterns or muscle rigidity, however, vocal detection could prevent the extent of these symptoms by diagnosing the patient early on. The question asked in this study is can we find significant biomarkers between people who do and do not have Parkinson's and can it be quantified from speech frequencies and phonetics.
Beshay, Rachelle; Gill, Madison; Shiraishi, Michael; and Soangra, Rahul, "Identifying Voice-based Digital Biomarkers of Parkinson's Disease" (2022). Student Scholar Symposium Abstracts and Posters. 566.