Publication Details
Title: Speech Recognition Experiments on Switchboard Corpus
Author: T. Abe
Group: ICSI Technical Reports
Date: March 2000
PDF: ftp://ftp.icsi.berkeley.edu/pub/techreports/2000/tr-00-004.pdf
Overview:
This report shows results of a set of speech recognition approaches performed on Switchboard corpus which is a large spontaneous telephone conversation database. The purpose is to improve recognition accuracy on Switchboard with our connectionist hybrid model. The methods include the choice of kinds of acoustic features, gender dependent training, use of multi-stream features, etc. We will also show that adding a feature of periodicity measure improves recognition accuracy. Finally, we will show a speaker adaptation approach that improves recognition accuracy for speech by a particular speaker.
Bibliographic Information:
ICSI Technical Report TR-00-004
Bibliographic Reference:
T. Abe. Speech Recognition Experiments on Switchboard Corpus. ICSI Technical Report TR-00-004, March 2000
Author: T. Abe
Group: ICSI Technical Reports
Date: March 2000
PDF: ftp://ftp.icsi.berkeley.edu/pub/techreports/2000/tr-00-004.pdf
Overview:
This report shows results of a set of speech recognition approaches performed on Switchboard corpus which is a large spontaneous telephone conversation database. The purpose is to improve recognition accuracy on Switchboard with our connectionist hybrid model. The methods include the choice of kinds of acoustic features, gender dependent training, use of multi-stream features, etc. We will also show that adding a feature of periodicity measure improves recognition accuracy. Finally, we will show a speaker adaptation approach that improves recognition accuracy for speech by a particular speaker.
Bibliographic Information:
ICSI Technical Report TR-00-004
Bibliographic Reference:
T. Abe. Speech Recognition Experiments on Switchboard Corpus. ICSI Technical Report TR-00-004, March 2000