Recurrent deep neural networks for robust speech recognition
Recurrent deep neural networks for robust speech recognition
复制标题
DOI：
10.1109/icassp.2014.6854661
复制
发表时间：
2014-05
期刊：
影响因子：
--
通讯作者：
Chao Weng;Dong Yu;Shinji Watanabe;B. Juang
中科院分区：
文献类型：
--
作者：
Chao Weng;Dong Yu;Shinji Watanabe;B. Juang
在这项工作中，我们提出了用于鲁棒自动语音识别（ASR）的循环深度神经网络（DNN）。完整的循环连接被添加到传统前馈 DNN 的某些隐藏层，并允许模型捕获深度表示中的时间依赖性。引入了一种新的随时间反向传播 (BPTT) 算法，使所提出的循环 DNN 上的小批量随机梯度下降 (SGD) 更加高效和有效。我们在第二次 CHiME 挑战（轨道 2）和 Aurora-4 任务的混合设置下评估了所提出的循环 DNN 架构。 CHiME 挑战数据的实验结果表明，与 DNN 系统相比，所提出的系统可以获得一致的 7% 相对 WER 改进，无需前端预处理、说话者自适应训练或多次解码即可实现最先进的性能。对于 Aurora-4 上的实验，所提出的系统比强大的 DNN 基线系统实现了 4% 的相对 WER 改进。
In this work, we propose recurrent deep neural networks (DNNs) for robust automatic speech recognition (ASR). Full recurrent connections are added to certain hidden layer of a conventional feedforward DNN and allow the model to capture the temporal dependency in deep representations. A new backpropagation through time (BPTT) algorithm is introduced to make the minibatch stochastic gradient descent (SGD) on the proposed recurrent DNNs more efficient and effective. We evaluate the proposed recurrent DNN architecture under the hybrid setup on both the 2nd CHiME challenge (track 2) and Aurora-4 tasks. Experimental results on the CHiME challenge data show that the proposed system can obtain consistent 7% relative WER improvements over the DNN systems, achieving state-of-the-art performance without front-end preprocessing, speaker adaptive training or multiple decoding passes. For the experiments on Aurora-4, the proposed system achieves 4% relative WER improvement over a strong DNN baseline system.