Advancing Speech Activity Detection for Automatic Speech Assessment of Pre-School Children Prompted Speech
MetadataShow full item record
Early screening of developmental disorders is important for the well-being of children and their families. Earlier screening leads to earlier diagnosis and intervention which can reduce the risk of long-term social, emotional, and academic problems. The American Academy of Pediatrics recommends developing standardized screening tools to support health care professionals, as well parents and caregivers in the process of early identification of developmental disorders. Speech sound disorder (SSD), a developmental disorder which affects children’s ability to produce the sounds of their native language, has a prevalence rate of 3-16% among children in the United States. Screening for SSDs generally requires professional collection, recording, evaluation, and decision-making by a certified speech-language pathologist. Automating part or all of this process could significantly reduce the amount of time and effort allocated to the screening process. However, in order for this process and especially the final ‘pass’ or ‘fail’ decision to be automated, children’s speech content must be extracted from silence or background noise that may be present while the speech content is being collected. In this thesis, we explore an existing unsupervised speech activity detection algorithm based on a combination of five robust speech-based features using Gaussian Mixture Models and modify the same to detect child speech.