In late 2009, we wrote about the Language Environment Analysis (LENA) system, which identifies autism in children as young as 18 months by labeling vocalizations from recordings and generating automatic acoustic analysis of those sounds.
At the time, LENA claimed its system was 91 percent accurate. Now, according to research published this week in the Proceedings of the National Academy of Sciences, that number has dropped slightly to 86 percent.
The researchers analyzed more than 3 million child utterances from 1,486 all-day recordings of 232 children. Of the 12 acoustic parameters analyzed for vocal development, syllabification (producing well-formed syllables … Read more