arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1601.05936cs.CLcs.LGstat.ML
收藏

利用低维结构增强语音识别中基于DNN的声学建模

Exploiting Low-dimensional Structures to Enhance DNN Based Acoustic Modeling in Speech Recognition

Idiap Research Institute(伊迪亚普研究所);University of Fribourg(弗里堡大学)

Pranay Dighe, Gil Luyet, Afsaneh Asaei, Herve Bourlard

摘要:

我们建议将深度神经网络(DNN)类条件后验概率的声学空间建模为低维子空间的联合。为此,训练后验用于字典学习和稀疏编码。使用该字典的测试后验的稀疏表示使得能够投影到训练数据的空间。依赖于这样的事实,即后验子空间的内在尺寸确实是非常小的,属于一类的所有后验的矩阵具有非常低的秩,我们演示了如何低维结构,使进一步增强后验和纠正虚假的错误,由于失配条件。增强的声学建模方法导致使用混合DNN-HMM(隐马尔可夫模型)框架在干净和噪声条件下的连续语音识别任务的改进,其中实现了高达15.4%的字出错率(WER)的相对降低。

英文摘要:

We propose to model the acoustic space of deep neural network (DNN) class-conditional posterior probabilities as a union of low-dimensional subspaces. To that end, the training posteriors are used for dictionary learning and sparse coding. Sparse representation of the test posteriors using this dictionary enables projection to the space of training data. Relying on the fact that the intrinsic dimensions of the posterior subspaces are indeed very small and the matrix of all posteriors belonging to a class has a very low rank, we demonstrate how low-dimensional structures enable further enhancement of the posteriors and rectify the spurious errors due to mismatch conditions. The enhanced acoustic modeling method leads to improvements in continuous speech recognition task using hybrid DNN-HMM (hidden Markov model) framework in both clean and noisy conditions, where upto 15.4% relative reduction in word error rate (WER) is achieved.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑