Robust ASR based on clean speech models: an evaluation of missing data techniques for connected digit recognition in noise
Robust ASR based on clean speech models: an evaluation of missing data techniques for connected digit recognition in noise
复制标题
基于干净语音模型的鲁棒 ASR：噪声中连接数字识别的缺失数据技术评估
DOI：
10.21437/eurospeech.2001-76
复制
发表时间：
2001
影响因子：
6.5
通讯作者：
P. Green
中科院分区：
文献类型：
--
作者：
J. Barker;M. Cooke;P. Green
在这项研究中，技术的分类与丢失或不可靠的数据适用于自动语音识别（ASR）的噪声鲁棒性的问题。所描述的技术对任何噪声背景做了最小的假设，而是依赖于对干净语音的了解。使用Aurora 2连接数字识别任务对系统进行评估。使用干净的语音训练的模型，我们获得了65%的相对改善极光干净的训练基线系统，与极光基线的多条件训练的性能相当。
In this study, techniques for classification with missing or unreliable data are applied to the problem of noise-robustness in Automatic Speech Recognition (ASR). The techniques described make minimal assumptions about any noise background and rely instead on what is known about clean speech. A system is evaluated using the Aurora 2 connected digit recognition task. Using models trained on clean speech we obtain a 65% relative improvement over the Aurora clean training baseline system, a performance comparable with the Aurora baseline for multicondition training.