|
181. |
Keyword and phrase spotting by use of the Harpy speech system |
|
The Journal of the Acoustical Society of America,
Volume 64,
Issue S1,
1978,
Page 182-182
Bruce T. Lowerre,
Raj Reddy,
Preview
|
PDF (183KB)
|
|
摘要:
Two problems of a keyword spotting system that are not encountered in isolated work recognition systems are the detection of the beginning and ending of the words within the context of other words and the phenomena that occur at the word junctures. A modification of the Harpy connected speech recognition system has yielded a word and phrase spotting system that deals with these problems. This word spotting system incorporates the lexical and word juncture knowledge and uses a modified beam search technique of the Harpy system. This enables the system to not only rapidly and accurately spot single words within any context but also multiple words and/or complex phrases. Performance results of the system will be presented for various words and phrases and for different threshold values.
ISSN:0001-4966
DOI:10.1121/1.2004067
出版商:Acoustical Society of America
年代:1978
数据来源: AIP
|
182. |
Feature selection using adaptive learning networks for text‐independent speaker verification |
|
The Journal of the Acoustical Society of America,
Volume 64,
Issue S1,
1978,
Page 183-183
R. S. Cheung,
Preview
|
PDF (156KB)
|
|
摘要:
A nonlinear transformation which makes use of the polynomial discriminant function (PDF) is applied to the selection of features for text‐independent speaker verification. For each speaker within the set, a PDF is established such that it maps the original attribute set onto the one‐dimensional one resulting in the best discrimination of the speaker from the others in the new feature space. The complicated PDF is implemented using the technique known as the adaptive learning network (ALN). In this case, the multinomial discriminant function is estimated using interconnecting fundamental building blocks each of which computes a two‐element quadratic discriminant function. Through a comprehensive training procedure, the coefficients needed to describe each building block and the interconnecting configurations of them are determined. Application of the chosen attributes to a text‐independent speaker verification experiment yields relatively low false speaker rejection and verification error rates.
ISSN:0001-4966
DOI:10.1121/1.2004072
出版商:Acoustical Society of America
年代:1978
数据来源: AIP
|
|