张威等：基于DL-T及迁移学习的语音识别研究 441 to-end speech recognition /Conference of the International networks.Chin Comput,2017,40(6):1229 Speech Communication Association.Graz,2019:4395 (周飞燕，金林鹏，董军.卷积神经网络研究综述.计算机学报， [16]Bu H,Du J Y,Na X Y,et al.Aishell-1:an open-source mandarin 2017,40(6):1229) speech corpus and a speech recognition baseline[J/OL].arYiv [23]Yi J Y,Tao J H,Liu B,et al.Transfer learning for acoustic preprint(2017-09-16)[2019-10-10].http://arxiv..org/abs/17- modeling of noise robust speech recognition.J Tsinghua Univ Sci 09.05522 Technol,,2018,58(1):55 [17]Battenberg E,Chen J T.Child R,et al.Exploring neural (易江燕，陶建华，刘斌，等.基于迁移学习的噪声鲁棒性语音识 transducers for end-to-end speech recognition //2017 IEEE 别声学建模.清华大学学报：自然科学版，2018.58(1)：55) Automatic Speech Recognition and Understanding Workshop [24]Xue J B,Han J Q,Zheng T R,et al.A multi-task learning (ASRU).Okinawa,2017:206 framework for overcoming the catastrophic forgetting in automatic [18]Williams R J,Zipser D.Gradient-based learning algorithms for speech recognition[J/OL].arXiv preprint (2019-04-17)[2019-10- recurrent networks and their computational complexity /Back- 101.https://arxiv.org/abs-/1904.08039 propagation:Theory,Architectures and Applications.1995:433 [25]Mikolov T,Sutskever I,Chen K,et al.Distributed representations [19]Huang G,Liu Z,Maaten L V D,et al.Densely connected of words and phrases and their compositionality /Proceedings of convolutional networks /IEEE Conference on Computer Vision the 26th International Conference on Neural Information and Pattern Recognition.Honolulu,2017:4700 [20]Cao Y,Huang Z L,Zhang W,et al.Urban sound event Processing Systems-Volume 2.Canada,2013:3111 classification with the N-order dense convolutional network.J [26]Povey D.Ghoshal A,Boulianne G,et al.The Kaldi speech Xidian Univ Nat Sci,2019,46(6):9 recognition toolkit IEEE 2011 Workshop on Automatic Speech (曹毅，黄子龙，张威，等.N-DenseNetl的城市声音事件分类模型 Recognition and Understanding.Big Island,2011 西安电子科技大学学报：自然科学版，2019,46(6)：9) [27]Paszke A,Gross S,Chintala S,et al.Automatic differentiation in [21]Zhang S,Gong Y H,Wang J J.The development of deep PyTorch 31st Conference on Neural Information Processing convolutional neural networks and its application in computer Systems.Long Beach,2017 vision.ChinJ Comput,2019,42(3):453 [28]Shan C,Weng C,Wang G,et al.Component fusion:learning (张顺，龚怡宏，王进军.深度卷积神经网络的发展及其在计算 replaceable language model component for end-to-end speech 机视觉领域的应用.计算机学报，2019,42(3)：453) recognition system IEEE International Conference on Acoustics. [22]Zhou F Y,Jin L P,Dong J.Review of convolutional neural Speech and Signal Processing.Brighton,2019:5361to-end  speech  recognition  // Conference of the International Speech Communication Association. Graz, 2019: 4395 Bu H, Du J Y, Na X Y, et al. Aishell-1: an open-source mandarin speech  corpus  and  a  speech  recognition  baseline[J/OL]. arXiv preprint (2017-09-16)[2019-10-10]. http://arxiv.org/abs/17- 09.05522 [16] Battenberg  E,  Chen  J  T,  Child  R,  et  al.  Exploring  neural transducers  for  end-to-end  speech  recognition  //  2017 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU). Okinawa, 2017: 206 [17] Williams  R  J,  Zipser  D.  Gradient-based  learning  algorithms  for recurrent  networks  and  their  computational  complexity  // Back￾propagation: Theory, Architectures and Applications. 1995: 433 [18] Huang  G,  Liu  Z,  Maaten  L  V  D,  et  al.  Densely  connected convolutional  networks  // IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, 2017: 4700 [19] Cao  Y,  Huang  Z  L,  Zhang  W,  et  al.  Urban  sound  event classification  with  the  N-order  dense  convolutional  network. J Xidian Univ Nat Sci, 2019, 46（6）: 9 （曹毅, 黄子龙, 张威, 等. N-DenseNet的城市声音事件分类模型. 西安电子科技大学学报: 自然科学版, 2019, 46（6）：9） [20] Zhang  S,  Gong  Y  H,  Wang  J  J.  The  development  of  deep convolutional  neural  networks  and  its  application  in  computer vision. Chin J Comput, 2019, 42（3）: 453 （张顺, 龚怡宏, 王进军. 深度卷积神经网络的发展及其在计算 机视觉领域的应用. 计算机学报, 2019, 42（3）：453） [21] [22] Zhou  F  Y,  Jin  L  P,  Dong  J.  Review  of  convolutional  neural networks. Chin J Comput, 2017, 40（6）: 1229 （周飞燕, 金林鹏, 董军. 卷积神经网络研究综述. 计算机学报, 2017, 40（6）：1229） Yi  J  Y,  Tao  J  H,  Liu  B,  et  al.  Transfer  learning  for  acoustic modeling of noise robust speech recognition. J Tsinghua Univ Sci Technol, 2018, 58（1）: 55 （易江燕, 陶建华, 刘斌, 等. 基于迁移学习的噪声鲁棒性语音识 别声学建模. 清华大学学报: 自然科学版, 2018, 58（1）：55） [23] Xue  J  B,  Han  J  Q,  Zheng  T  R,  et  al.  A  multi-task  learning framework for overcoming the catastrophic forgetting in automatic speech  recognition[J/OL]. arXiv preprint (2019-04-17)[2019-10- 10]. https://arxiv.org/abs-/1904.08039 [24] Mikolov T, Sutskever I, Chen K, et al. Distributed representations of words and phrases and their compositionality // Proceedings of the 26th International Conference on Neural Information Processing Systems - Volume 2.Canada, 2013: 3111 [25] Povey  D,  Ghoshal  A,  Boulianne  G,  et  al.  The  Kaldi  speech recognition  toolkit  // IEEE 2011 Workshop on Automatic Speech Recognition and Understanding. Big Island, 2011 [26] Paszke A, Gross S, Chintala S, et al. Automatic differentiation in PyTorch  //  31st Conference on Neural Information Processing Systems. Long Beach, 2017 [27] Shan  C,  Weng  C,  Wang  G,  et  al.  Component  fusion:  learning replaceable  language  model  component  for  end-to-end  speech recognition system // IEEE International Conference on Acoustics, Speech and Signal Processing. Brighton, 2019: 5361 [28] 张    威等： 基于 DL-T 及迁移学习的语音识别研究 · 441 ·