Speaker Change Detection and Retrieval for Automatic Minute Taking

As speech recognition of broadcast news is becoming
more mature, research is moving into types of speech
that are more challenging. One such area is
conversational speech. Initially telephone
conversations were studied but more recently
attention moved to meeting recordings. Meetings are
social events, were people exchange information.

This book has presented a study on an automatic
minute tacking (AMT) system. By the use of particular
characteristic provided by SVM classification, this
book successfully proposes the SVM training
misclassification rate (STMR) algorithm for the
application of speaker change detection (SCD).
Experimental results demonstrate that the STMR
performance of outperforms that of conventional
methods. Also, this book proposes two partial
matching algorithms and a keyword matching algorithm
that can be used to retrieve spoken sentences both on
training free and language independent.

Autorentext
Po-Chuan Lin was born in Taipei, Taiwan, 1976. He received the Phd degree in electrical engineering from the National Cheng- Kung University, Tainan, Taiwan, in 2007. His research interests include speech signal processing, VLSI architecture design and embedded system design.

Klappentext
As speech recognition of broadcast news is becoming more mature, research is moving into types of speech that are more challenging. One such area is conversational speech. Initially telephone conversations were studied but more recently attention moved to meeting recordings. Meetings are social events, were people exchange information. This book has presented a study on an automatic minute tacking (AMT) system. By the use of particular characteristic provided by SVM classification, this book successfully proposes the SVM training misclassification rate (STMR) algorithm for the application of speaker change detection (SCD). Experimental results demonstrate that the STMR performance of outperforms that of conventional methods. Also, this book proposes two partial matching algorithms and a keyword matching algorithm that can be used to retrieve spoken sentences both on training free and language independent.