CASIA OpenIR  > 模式识别国家重点实验室  > 语音交互
易江燕1,2; 陶建华1,2,3; 刘斌1; 温正棋1
Source Publication清华大学学报
Other AbstractIn order to improve the performance of robust speech recognition under noisy environments, this paper proposes an approach to train acoustic models using transfer learning. This method is that the training of an acoustic model trained with noisy data (student model) is guided by an acoustic model trained with clean data (teacher model). Such training process forces the posterior probability distribution of the student model to be close to the teacher model. Thus it can be achieved by minimizing the Kullback-Leibler (KL) Divergence between the posterior probability distribution of the student model and the teacher model. Experimental results on CHiME-2 dataset show that the proposed method achieves 7.29% absolutely average word error rate (WER) improvement over the baseline and achieves 3.92% absolutely average WER improvement over the best system of CHiME-2.
Keyword鲁棒语音识别 声学模型 神经网络 迁移学习
Indexed ByEI
Document Type期刊论文
Corresponding Author陶建华
Recommended Citation
GB/T 7714
易江燕,陶建华,刘斌,等. 基于迁移学习的鲁棒语音识别声学建模方法[J]. 清华大学学报,2018,58(1):55-60.
APA 易江燕,陶建华,刘斌,&温正棋.(2018).基于迁移学习的鲁棒语音识别声学建模方法.清华大学学报,58(1),55-60.
MLA 易江燕,et al."基于迁移学习的鲁棒语音识别声学建模方法".清华大学学报 58.1(2018):55-60.
Files in This Item: Download All
File Name/Size DocType Version Access License
基于迁移学习的噪声鲁棒语音识别声学建模.(1412KB)期刊论文作者接受稿开放获取CC BY-NC-SAView Download
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[易江燕]'s Articles
[陶建华]'s Articles
[刘斌]'s Articles
Baidu academic
Similar articles in Baidu academic
[易江燕]'s Articles
[陶建华]'s Articles
[刘斌]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[易江燕]'s Articles
[陶建华]'s Articles
[刘斌]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: 基于迁移学习的噪声鲁棒语音识别声学建模.pdf
Format: Adobe PDF
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.