Review on Dysarthric Speech Severity Level Classification Frameworks

Dysarthria is a motor speech disorder that results from malfunctions in the mechanisms and systems which regulate the movements required for speech production. It is challenging to organise, programme, manage, coordinate, and carry out speech productions due to these neurological limitations. Assessing the degree of dysarthria severity can provide information about the patient’s progress, assist pathologists in planning a treatment plan, and make automated voice recognition software for dysarthria more useful. This paper provides a thorough analysis of fresh and cutting-edge machine learning and deep learning methods and algorithms for dysarthric speech severity level classification. Depending on the degree of the impairment, various acoustic features and feature selection techniques are considered for the precise classification of dysarthric speech. This study is focused on various deep learning methodologies as well as the comparison of outcomes with various machine learning algorithms. An overview of the data sets and evaluation metrics frequently utilised in the classification of dysarthric severity levels is also included in this review. Finally, an audio-video cross model approach is reviewed for the comparison of severity level classification based on different features using different classifiers.