A New Method to Analyzing Speech to Detect Patients by Using Machine Learning Algorithms

The progressive death of dopaminergic neurons in the substantia nigra causes motor system failure in the common neurodegenerative illness known as chronic degenerative disorder (CDD). Researchers have looked at a wide variety of signals, including electroencephalography (EEG), gait analysis, and speech analysis, in an effort to diagnose chronic degenerative disorder (CDD). For the vast majority (over 95%) of people with Chronic degenerative disorder (CDD), the study of speech is the primary technique utilized to improve speech-related issues. This research introduces the use of voice analysis as a fresh approach to the diagnosis of chronic degenerative disorders. In the first step, we use a Stochastic optimization (SO) to select the best features from among the collected data. Then, to separate healthy people from those with Chronic degenerative condition, a network based on support vector machines (SVMs) is employed. The dataset employed in this investigation is comprised of biological speech signals from 38 participants, including 32 with a chronic degenerative condition diagnosis and 9 healthy controls.

The subjects were asked to say the letter "A" out loud for a full four seconds. Extracted from the signals were a total of 29 features, both linear and non-linear properties. F0 (the fundamental frequency), jitter, and the noise-to-harmonics ratio were the focus of analysis for 15 of the aforementioned characteristics. These components have long been acknowledged for their crucial roles in shaping speech signals' defining features. Because of the obvious effect these qualities have on people with Chronic degenerative illness, a careful selection of the best possible qualities was made. The data was divided into groups based on how many ideal features they had. Using these 5 improved features, we were able to achieve a 95.53% accuracy rate in our classifications.