{exciting.research} Elephantese

If you would ever wonder what it would take to understand the elephant you came across in the zoo (or perhaps in a rare street festival or a circus), worry not any more, because computers can translate Elephantese for us!

BIOACOUSTICS SOFTWARE DETECTS INDIVIDUAL ELEPHANT SPEECH PATTERNS
Hidden Markov model (HMM) algorithm-based PCs can interpret up to 85 percent of "elephantese," or elephant-talk, according to an article from Computer Magazine. A bioacoustics team from Marquette University, in conjunction with researchers from Disney's Animal Kingdom, say elephants, whose vocal range is 7 to 200 Hz, produce simultaneous rumbles in herds that are inaudible to humans. A trained HMM program extracted distinct elephant vocalizations through a sequence of "distilled data reductions, "which were strung into a chain to study new sound sequences, or individual elephant voices, for social context, language, and grammar. Once the program was trained to differentiate between voices, it could produce a record of individual elephant vocalizations. Read more in the article "What Jumbo Tells Dumbo".

Source: WHAT'S NEW @ IEEE IN COMPUTING
VOLUME 6 NUMBER 9 SEPTEMBER 2005