DocumentCode
302324
Title
Automatic recognition of Danish natural numbers for telephone applications
Author
Jacobsen, Claus N. ; Wilpon, Jay G.
Author_Institution
AT&T Bell Labs., Murray Hill, NJ, USA
Volume
1
fYear
1996
fDate
7-10 May 1996
Firstpage
459
Abstract
In major areas of the world, it is custom to cluster the digits of long numbers into groups of two or three digits and pronounce them using natural numbers such as “eighty/five” rather than the digits `eight five´. Forcing the users of automatic services to change their speaking styles could seriously limit the acceptance of such services. So far, however, it has not been considered possible to obtain sufficiently high accuracies with natural numbers. In this paper, we present a series of experiments that improved our string accuracy to 93.3% for Danish 8-digit telephone numbers. This is very close to the accuracy obtained for connected digits, 95.3%, in the same language
Keywords
natural languages; speech recognition; telephony; Danish 8-digit telephone numbers; Danish natural numbers; automatic services; automatic speech recognition; connected digits recognition; experiments; string accuracy; telephone applications; Automatic speech recognition; Cepstral analysis; Context modeling; Hidden Markov models; Jacobian matrices; Niobium; Pattern matching; Signal analysis; Speech analysis; Telephony;
fLanguage
English
Publisher
ieee
Conference_Titel
Acoustics, Speech, and Signal Processing, 1996. ICASSP-96. Conference Proceedings., 1996 IEEE International Conference on
Conference_Location
Atlanta, GA
ISSN
1520-6149
Print_ISBN
0-7803-3192-3
Type
conf
DOI
10.1109/ICASSP.1996.541132
Filename
541132
Link To Document