Event Based System: A Probabilistic Acoustic-Phonetic Approach to Automatic Speech Recognition (ISR IP)

For more information, contact ISR External Relations Director
Jeff Coriale at coriale@umd.edu or 301.405.6604.

ISR intellectual property available to license

Inventors: Carol Espy-Wilson, Amit Juneja

Description
The state of the art voice recognition systems use very little knowledge of acoustic-phonetics and are largely dependent on statistical modeling.

This invention, deemed a substantial advancement in the art, brings together knowledge of acoustic-phonetics and statistical pattern recognition to build a system, called Event-Based System (EBS) for automatic speech recognition. The system is based on representation of speech sounds by bundles of binary valued phonetic features.

This system not only obtains better accuracy in identification but also, unlike other systems, does not require initial training to 'learn" to recognize.

For more information
If you would like to license this intellectual property, have questions, would like to contact the inventors, or need more information, contact ISR External Relations Director Jeff Coriale at coriale@umd.edu or 301.405.6604.

Find more ISR IP
You can go to our main IP search page to search by research category or faculty name. Or view the entire list of available IP on our complete IP listing page.

ISR-IP-Espy-Wilson ISR-IP-speech

Published June 21, 2007