A Transformable Mel-Spectrogram for Diagnosis of Dysarthria Using a CNN Model

In recent years, brain diseases have become more common each year. This is because the population is growing and people are living longer.Some AI-based systems cannot work right when people have speech problems caused by neurological speech disorders, psychological speech disorders, or communication disorders. The disorder known as dysarthria is a neurological speech disease that makes it difficult for people to speak clearly. Tracking the status of these individuals with dysarthria over time might be challenging if they are not identified and treated promptly. As the patient’s symptoms worsen, they can affect their mental and physical well-being. A large body of prior work on dysarthria has included categorization algorithms based on machine learning or deep learning. The proposed study gets better results on the TORGO dataset by turning the audio samples into Mel spectrograms and using a convolutional neural network model based on transfer learning. The proposed work performed with an accuracy of 99.5%.