Improving speech recognition in reverberation using a room-aware deep neural network and multi-task learning
Improving speech recognition in reverberation using a room-aware deep neural network and multi-task learning
复制标题
DOI：
10.1109/icassp.2015.7178925
复制
发表时间：
2015-04
期刊：
影响因子：
--
通讯作者：
Ritwik Giri;M. Seltzer;J. Droppo;Dong Yu
中科院分区：
文献类型：
--
作者：
Ritwik Giri;M. Seltzer;J. Droppo;Dong Yu
在本文中，我们提出了两种方法来改进用于混响环境中语音识别的深度神经网络（DNN）声学模型。两种方法都在训练 DNN 时利用辅助信息，但信息类型及其使用方式有所不同。第一种方法使用并行训练数据进行多任务学习，其中网络被训练以使用共享表示来执行主要语义分类任务和次要特征增强任务。第二种方法使用从观测信号中提取的混响环境参数化来训练房间感知 DNN。在 REVERB Challenge 语料库的单麦克风任务上进行了实验。该方法在 SimData 测试集上获得了 7.8% 的字错误率，低于使用相同训练数据和评估条件的所有报告系统，在不匹配的 RealData 测试集上获得了 27.5% 的字错误率，低于除两个系统之外的所有系统。
In this paper, we propose two approaches to improve deep neural network (DNN) acoustic models for speech recognition in reverberant environments. Both methods utilize auxiliary information in training the DNN but differ in the type of information and the manner in which it is used. The first method uses parallel training data for multi-task learning, in which the network is trained to perform both a primary senone classification task and a secondary feature enhancement task using a shared representation. The second method uses a parameterization of the reverberant environment extracted from the observed signal to train a room-aware DNN. Experiments were performed on the single microphone task of the REVERB Challenge corpus. The proposed approach obtained a word error rate of 7.8% on the SimData test set, which is lower than all reported systems using the same training data and evaluation conditions, and 27.5% on the mismatched RealData test set, which is lower than all but two systems.