High-frequency components do not enhance computer recognition of dysarthric speech

Main Category: Medical Devices / Diagnostics
Article Date: 27 Jul 2005 - 5:00 PDT

email to a friend   printer friendly   opinions  

This study describes the development of an artificially intelligent communication/control tool for speech and motor-impaired individuals.

A hidden Markov model (HMM)-based system for recognition of dysarthric speech was constructed, with special emphasis on the effect of high-frequency spectral components on the system's recognition rate.

A set of 15 words and 10 digits were recorded from three male cerebral palsy patients with moderate dysarthria.

Investigators found that inclusion of high-frequency components provided no significant useful information that would enhance speech recognition.

The application of speech recognition technology to impaired speech would enable people with such impairments to control mobility and appliances.

High-frequency components do not enhance computer recognition of dysarthric speech, pg. 363

About the Journal JRRD (Journal of Rehabilitation Research and Development) has been a leading research journal in the field of rehabilitation medicine and technology for more than 40 years. JRRD, a peer-reviewed, scientifically indexed journal, publishes original research papers, review articles, as well as clinical and technical commentary from U.S. and international researchers on all rehabilitation research disciplines. JRRD's mission is to responsibly evaluate and disseminate scientific research findings impacting the rehabilitative healthcare community. For more information about JRRD, visit http://www.vard.org.

The current issue of the Journal of Rehabilitation Research and Development (JRRD) includes articles that focus on interpersonal aspects of stroke rehabilitation, depression following stroke, stress and spinal cord injury, detection and treatment of diabetic ulcers, and sensory and communication evaluation tools. Full-text articles are available, free, online at http://www.vard.org.

Judith LaVoie
judith@vard.org
410-962-1800 x229
VA Research Communications Service

• Additional
• References
• Citations