Robust scoring of voice exercises in computer-based speech therapy systems

Abstract

Speech therapy is essential to help children with speech sound disorders. While some computer tools for speech therapy have been proposed, most focus on articulation disorders. Another important aspect of speech therapy is voice quality but not much research has been developed on this issue. As a contribution to fill this gap, we propose a robust scoring model for voice exercises often used in speech therapy sessions, namely the sustained vowel and the increasing/decreasing pitch variation exercises. The models are learned with a support vector machine and double cross-validation, and obtained accuracies from approximately 73.98% to 85.93% while showing a low rate of false negatives. The learned models allow classifying the children's answers on the exercises, thus providing them with real-time feedback on their performance.