Synthetic Data Generation Techniques for Developing AI-based Speech Assessments for Parkinson's Disease (A Comparative Study)

Voice impairments are common problems for patients with Parkinson's disease (PD) [1] and other types of dementia [2]. Many PD patients face difficulties in speech and language generation, particularly in modulating and articulating their voices or generating and inflecting verbs [3, 4]. Accurately diagnosing language impairments associated with PD can significantly assist clinicians in identifying individuals with PD [3, 4, 5]. Current advances in artificial intelligence (AI) [5, 2, 6] have motivated clinicians to leverage AI-based speech assessments to detect speech assessments associated with PD and identify patients with PD. These AI should be developed based on machine learning (ML) or deep learning classifiers, which are trained using extracted acoustic and linguistic features from patients' voices.