Mapping of Speech Front-End Signal Processing to High Performance Vector Architectures

TitleMapping of Speech Front-End Signal Processing to High Performance Vector Architectures
Publication TypeTechnical Report
Year of Publication1995
AuthorsMoretto, P.
Other Numbers1003
Abstract

Front-end signal processing is a crucial stage for speech recognition systems. The capability of operating in adverse conditions, with high background noise and different channel characteristics, is one of the major goals when developing automatic speech recognition systems for use in real world environments. We describe the study of the mapping of a fundamental part of speech recognition systems - a robust speech front end algorithm called RASTA - to the Torrent architecture. The mapping problem is particularly relevant because at the moment there is no efficient automatic tool for implementing algorithms on this architecture.

URLhttp://www.icsi.berkeley.edu/ftp/global/pub/techreports/1995/tr-95-063.pdf
Bibliographic Notes

ICSI Technical Report TR-95-063

Abbreviated Authors

P. Moretto

ICSI Publication Type

Technical Report