Machine Learning in the Identification, Classification, and Treatment of Speech Sound Disorders: A Survey

A speech sound disorder (SSD) is a speech disorder in which children cannot produce or use some sounds (phonemes) past the expected age. SSD is the world’s most common type of communication disorder, and it affects approximately 10% to 15% of preschoolers and 6% of students. In the past few decades, researchers focused on how to identify and design the right treatment methods. Machine learning has been widely used in solving problems from various domains. Researchers investigated how to use machine learning approaches in identification, classification, and treatment of SSD in a more effective and efficient way. A survey study on this area can give some new insights to researchers further investigation, so there does not exist a literature review in this area. This survey can fill this gap. In this survey, we discuss the form of data for analysis, methods of data augmentation, adopted feature extraction algorithms, prediction algorithms, performance evaluation techniques. This paper provides a structured view of the research on the use of machine learning in the field of speech sound disorders to date.