Acoustic adaptation to dynamic background conditions with asynchronous transformations
详细信息 查看全文
- 作者：Oscar Saz ; o.saztorralba@sheffield.ac.uk" class="auth_mail" title="E-mail the corresponding author ; Thomas Hain
- 关键词：Speech recognition ; Acoustic adaptation ; Factorisation ; Dynamic background ; Media transcription
- 刊名：Computer Speech & Language
- 出版年：2017
- 出版时间：January 2017
- 年：2017
- 卷：41
- 期：Complete
- 页码：180-194
- 全文大小：798 K
文摘
- class="listitem" id="list_ulist0010">
- class="label">•
We propose asynchronous CMLLR adaptation for complex backgrounds.
- class="label">•
Noise adaptive techniques and speaker factorisation are also explored.
- class="label">•
Results in a noisy WSJ benchmark task show up to 40% WER reduction.
- class="label">•
Evaluation in the transcription of multi-media data achieves 3% WER reduction.