Enhancing the Diagnosis of Speech Disorders: An In-Depth Investigation into Dysarthria Classification Using the ResNet18 Model

This research article concentrates on dysarthria, a speech impediment commonly observed in individuals with conditions such as cerebral palsy (CP) or amyotrophic lateral sclerosis (ALS). Early detection of dysarthria is essential for effective interventions and improved patient outcomes, as it significantly impacts communication abilities. Employing the TORGO database, comprising 2000 samples from individuals with and without dysarthria, encompassing diverse genders, the study utilizes a Convolutional Neural Network (ResNet18 Model) for speech classification. The dataset encompasses both dysarthric and non-dysarthric subjects of both genders, with 500 samples each, recorded in separate sessions. The ResNet18 Model-based approach demonstrates an impressive 96% accuracy in distinguishing between dysarthric and non-dysarthric instances, underscoring the potential of deep learning technology in aiding early dysarthria detection and facilitating timely interventions. This investigation significantly contributes to advancing diagnostic capabilities and presents a promising avenue to enhance the quality of life for individuals grappling with speech difficulties.