The goal of this project is to develop an Early Vocalization Analyzer (EVA), a computer program to automatically analyze recorded sample of infant vocalizations, and to begin building a normative data base of such vocalizations. This program will be the basis of a screening test, for use in hospitals and clinics, to evaluate which infants are at risk for later communication or other developmental problems. It may also be used to gather acoustic information on larger populations than has been previously possible. The major technical research goals of this phase are: classifying syllables by presence and position of consonants, identified by speech landmarks; identifying reduplicated sequences of such syllables (canonical babbling); classifying consonants by manner of production (nasal, fricative, etc.); classifying vowels as central versus peripheral. EVA will capture the spectral and temporal features that human judges use to classify babbles. This will permit reports of rate, development stage, and other attributes of an infants~ vocalizations. EVA will be tested on monthly recordings of infants from six to thirteen months.
EVA, a software screening and diagnostic tool that analyzes recordings of infant pre-speech vocalizations to determine if infants are at-risk for later communication difficulties, has the following commercial applications: it can be used by hospital infant follow-up clinics and early intervention centers to screen infants who could benefit from further diagnostic work up or therapeutic intervention; it can be used as a research tool to increase the knowledge-base of information on infant pre-speech vocalizations, particularly to establish a normative data base on typically developing children against which the performance of children from particular at-risk groups can be compared. The technology developed will also be the base of a system DIBS) to the stimulate babbling of infants with hearing disabilities.
