Evolving Diagnostic Techniques for Speech Disorders: Investigating Dysarthria Classification Through DenseNet201 CNN Framework

The primary subject of this research work pertains to dysarthria, a prevalent speech impairment observed in individuals diagnosed with cerebral palsy (CP) or amyotrophic lateral sclerosis (ALS). Prompt detection of dysarthria is essential for effective therapies and improved patient results, as it significantly impacts communication proficiency. The research employed a DenseNet201 Convolutional Neural Network (CNN) to categorize speech using the TORGO database, which has 2000 samples of individuals with and without dysarthria, representing various genders. The dataset comprises individuals of both genders, encompassing both dysarthric and non-dysarthric individuals. Each dataset consists of 500 samples, which were collected during distinct sessions. The DenseNet201 convolutional neural network (CNN) technique has a notable accuracy rate of 96% in distinguishing between cases of dysarthria and non-dysarthria. This outcome underscores the potential of deep learning technology in aiding the timely identification of dysarthria and facilitating fast interventions for affected individuals. This work provides valuable insights on the progress of diagnostic capabilities and offers a potential avenue for enhancing the well-being of those experiencing speech impairments.