Engineering Innovation Podcast and Radio Series

Clearer Conversation

PostedAugust 20, 2006

Download File (mp3)

Randy Atkins:  Your phone voice doesn’t sound entirely natural, because current infrastructure only allows transmission of a narrow sound wave range.
Harsha Sathyendra:  The problems occur mostly with the high frequency consonants like, N and M.  It’s very hard to distinguish let’s say “Nanny” and “Manny.”

Randy Atkins:  So Harsha Sathyendra, and other University of Florida engineering grad students, developed computer programs that guess intended sounds and reconstruct clipped wavelengths for the listener.  They tested it on other college students.

Harsha Sathyendra:  Ninety-three point one percent of the people preferred our reconstructed speech.
Randy Atkins:  With the National Academy of Engineering, Randy Atkins, WTOP Radio.