Dysarthric-speech detection using transfer learning with convolutional neural networks

Abstract

Speech Dysarthria is a disorder in which speech muscles become weak, and it becomes difficult to articulate otherwise linguistically normal speech. This work is based on detection of speech dysarthria and how it can assist physicians, specialists, and doctors in its detection. The proposed work achieves higher accuracies on the TORGO dataset by using a transfer learning based convolutional neural network model (TL-CNN) and by converting the audio samples to Mel-spectrograms. The proposed work TL-CNN achieved better accuracy when compared with other machine learning models.

Keywords

• Dysarthria
• Deep learning
• Medical technology
• Transfer learning
• Machine learning