Home // INTELLI 2012, The First International Conference on Intelligent Systems and Applications // View article

Swedish Speech Recognition Using Linear Time Normalization and Feature Selection Optimization

Authors:
Mattias Wahde

Keywords: Speech recognition; optimization; linear time normalization

Abstract:
A system for Swedish isolated word recognition has been developed, intended for use in an intelligent news reader agent for elderly care. The system uses linear time normalization of feature time series, as well as optimization (with a genetic algorithm) of both feature selection and feature weighting. The optimization of feature selection results in a potentially important decrease in the time needed to recognize a spoken word, while maintaining speech recognition performance.

Pages: 1 to 6

Copyright: Copyright (c) IARIA, 2012

Publication date: April 29, 2012

Published in: conference

ISSN: 2308-4065

ISBN: 978-1-61208-224-0

Location: Chamonix, France

Dates: from April 29, 2012 to May 4, 2012