[Colloq] PhD seminar

[Agnes Chan]

Please note the following changes in the PhD Seminars:

1/29/02:  Fell, "Early Vocalization, Analysis and Feeback"
          (see abstract below)
2/05/02:  NOTE time change 4:30 - 5:30
          Cooperman, "Issues in Parallel Computing"
2/19/02:  Attie

==================================================================
Early Vocalization, Analysis and Feedback

Professor Harriet Fell
Abstract

The Early Vocalization Analyzer (EVA), is a computer program that
automatically analyzes digitized recordings of infant vocalizations.
Using the landmark detection theory of Stevens et al for the recognition
of features in adult speech, EVA detects syllables in vocalizations
produced by typically infants. Landmarks are grouped into standard
syllable patterns and syllables are grouped into utterances. Principal
component analysis is used on this information to derive a "vocalization
age" that can clinically distinguish typically developing from
non-typically developing infants.

In a new project, we will develop sound recognition software to provide
real-time visual reinforcement of vocalizations produced by infants or
preschool children. This program is intended for children who are delayed
in the frequency and variety of pre-speech vocalizations they produce and,
hence, are at risk for being non-speaking. This will provide an automated
training system to maintain and develop children's motor plans necessary
for continued vocal development and expand the child's repertoire of
voluntary communicative behaviors that can be interpreted by partners in
functional
and social interaction. The system will encourage infants to practice
vocalizing and to explore a larger repertoire of phonetic patterns.
Particularly, it will reinforce the production of syllabic utterances
that are associated with later language and cognitive development.