Carnegie Mellon, University of Karlsruhe demonstrate breakthroughs in speech-to-speech translation


PITTSBURGH--Carnegie Mellon University and the University of Karlsruhe's joint International Center for Advanced Communication Technologies (InterACT) will hold an international videoconference at 9 a.m., Thursday, Oct. 27, to demonstrate new breakthroughs in cross-lingual communication.

The videoconference will take place in room 3305 Newell-Simon Hall on the Carnegie Mellon campus, and simultaneously at the University of Karlsruhe.

InterACT director, computer science professor Alex Waibel, who is a faculty member at both institutions, will demonstrate domain-independent, speech-to-speech translation in a lecture, which will be simultaneously translated from English to Spanish to German. According to Waibel, current speech-to-speech translation systems allow translation of spontaneous speech in very limited situations, like making hotel reservations or tourist shopping, but they cannot enable translation of large, open domains like lectures, television broadcasts, meetings or telephone conversations. The new technology developed by InterACT researchers fills that gap and makes it possible to extend such systems to other languages and lecture types.

Waibel also will illustrate new ways of delivering speech translation services beyond traditional headsets and an audio system. One involves an array of small ultra-sound speakers that can deliver a narrow beam of audio in a foreign language to a particular individual, while others nearby hear the same speech in the original language as it's spoken without disturbance.

Foreign language translation also can be produced through a system that tracks and measures electrical currents on the surface of a person's cheek and throat as they mouth words instead of speaking aloud. The system takes the signal off of electrodes that recognize muscle movement, translates and delivers I as audible sound in another language.

"Thus," said Waibel, "by moving our articulators in English, we can demonstrate the generation of speech in Spanish, German or other languages. In the future, such transducers could be implanted, enabling a speaker to produce any language at will."

In addition to these new technological breakthroughs, Waibel and his colleagues will also demonstrate other developments, including delivery of speech via heads-up display and text, a PDA-based pocket interpreter for fieldwork such as medical relief or military operations, as well as simultaneous translation of videos of European Parliamentary sessions.

Source: Eurekalert & others

Last reviewed: By John M. Grohol, Psy.D. on 21 Feb 2009
    Published on All rights reserved.