Blending Situation Awareness with Machine Learning to Identify Children’s Speech Disorders

Situation Awareness (SA) involves the correct interpretation of scenarios, allowing a system to respond to the observed environment and providing decision-making support in several domains. Speech therapy is a domain where SA may provide benefits; however, there are few proposals that address reasoning about situations to improve therapeutic tasks. An early identification of speech sound disorders allows the diagnosis and treatment of various pathologies. So, in this paper, we present a novel situation-aware approach using machine learning for detecting patterns on sound frequencies, aiming to classify the correctness in the pronunciation of words spoken by children aged 3 to 8 years. The approach was evaluated through a speech corpus containing approximately 27,000 audio files, collected from pronunciation assessments performed by Speech-Language Pathologists with more than 1,300 children. Our results showed an average accuracy over 92% for classifying speech disorders. The classification results were used to elaborate a projection strategy based on the identified situation.