Prediction of Speech Delay from Acoustic Measurements

Speech delay is characterized by a difficulty with producing or perceiving the sounds of language in comparison to one’s peers. It is a common problem in young children, occurring at a rate of about 5%. There are high rates of co-occurring problems with language, reading, learning, and social interactions, so intervention is needed for most. The GoldmanFristoe Test of Articulation (GFTA) is a standardized tool for the assessment of consonant articulation in American English children. GFTA scores are normalized for age and can be used to help diagnose and assess speech delay. The GFTA was administered to 65 young children, a mixture of delayed children and controls. Their productions of the 39 GFTA words spoken in isolation were recorded and aligned to 3-state hidden Markov models. Seven measurements (state log likelihoods, state durations, and total duration) were extracted from each target segment in each word. From a subset of these measures, cross-validated statistical models were used to predict the children's GFTA scores and whether they were delayed. The measurements most useful for prediction came primarily from approximants /r, l/. An analysis of the predictors and discussion of the implications will be provided.