Kids' Speech v1.1
The Kids' Speech Corpus was developed to facilitate research about
the characteristics of kids' speech at different ages and to train and
evaluate recognizers for use in language training and other
interactive tasks involving children. For instance, this corpus was
used to train recognizers used in
language development with deaf children.
In cooperation with the
Forest Grove School District
speech was gathered from children in grades K through 10. Approximately 100 children at each grade level
have been recorded.
The data collection was performed using the
CSLU Speech Toolkit and two Pentium Pro computers
running Windows NT. The protocol consists of a series of words and
sentences that the child was prompted to repeat by a computer animated
talking head. Each computer was manned by a CSLU staff member who
monitored progress and helped the child with any difficulties. The
average time spent by a child at the computer was 20 minutes, yielding
approximately 8-10 minutes of speech (16 bit, 16khz, mono). The data
are recorded through Soundblaster audio cards, with head-mounted
The following table shows the number of kids recorded for each grade
at the end of the collection.
The development of a protocol for this
data collection was driven by a variety of important principles. We
wanted to collect words, phrases, and fluent speech in a manner that
could be repeated for all of the children, regardless of age. This
necessitated words and phrases that were simple enough to be mimicked
by the youngest children (ages 5 or 6). In addition to this
simplicity requirement, we tried to get a sampling of the most common
biphones in as many contexts as possible.
Download a Sample
A small sample of the kids speech corpora may be downloaded for free: Kids'