Using Novel Hybrid Convolutional Neural Network for Dysarthria Diagnosis

Dysarthria is a motor speech disorder characterized by articulation and phonation difficulties resulting from speech muscle weakness, paralysis, or incoordination. A precise and timely diagnosis of dysarthria is essential for effective treatment and management of the condition, as it may deteriorate over time or be a precursor to a much more serious disease. On the other hand, this is becoming a severe problem in recent times owing to the rising ageing population and the prevalence of neurological disorders among such people. Hence early detection of dysarthria is deemed essential for the timely management of the disease. In recent years, Artificial Intelligence (AI) applications have shown promising results in various audio processing tasks and incorporated into pathological voice analysis for disease diagnosis. The majority of previous studies on dysarthria detection employed Machine Learning (ML) and Deep Learning (DL) models as the disease classification models. In light of this, this study presents a novel hybrid approach for classifying dysarthria based on audio data using Convolutional Neural Network (CNN) and Support Vector Machine (SVM). According to the experimental results, the proposed classification schema achieves an accuracy of 98.25 % compared to previous research work. Overall, our proposed method aims to automate the classification process, enabling faster and more reliable diagnoses.