A Novel Speech Recognition Approach in Non-stationary and Chaotic Environment

Period of Performance: 07/01/2004 - 04/30/2005

$70K

Phase 1 STTR

Recipient Firm

Intelligent Automation, Inc.
15400 Calhoun Dr, Suite 190
Rockville, MD 20855
Principal Investigator
Firm POC

Research Institution

Carnegie Mellon University
5000 Forbes Avenue
Pittsburgh, PA 15213
Institution POC

Abstract

Existing speech recognition software such as IBM via Voice works well in quiet and stationary background noise environments. However, the recognition performance drops quite significantly in crowded and noisy control room, battle stations, emergency room, factory floor, etc. The main reason is that the noise is non-stationary and chaotic. In this proposal, Intelligent Automation, Inc. (IAI) and its subcontractor, Prof. Richard Stern of Carnegie Mellon University (CMU), propose a novel system to improve the speech recognition performance in chaotic and non-stationary environment. The core technology will be a Robust Speech Recognition method developed by CMU. However, we will make an important improvement. The key idea is to use a new sensor called General Electromagnetic Movement Sensor (GEMS), which can be attached to the neck, to identify voiced and un-voiced regions in the speech. GEMS was designed and built by Aliph. IAI purchased one GEMS and used it for an Army project on multi-modal speech enhancement project. The approach combines the state-of-the-art technology in speech recognition and will significantly improve the recognition rate in non-stationary and chaotic environment.