arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2003.12687cs.CLcs.SDeess.AS
收藏

端到端重叠语音识别的串行输出训练

Serialized Output Training for End-to-End Overlapped Speech Recognition

Microsoft Corp.(微软公司)

Naoyuki Kanda, Yashesh Gaur, Xiaofei Wang, Zhong Meng, Takuya Yoshioka

摘要:

本文提出了一种基于注意力的编码器-解码器方法的多说话人重叠语音识别框架--串行输出训练(SOT)。与置换不变训练(PIT)不同,SOT使用只有一个输出层的模型,该模型一个接一个地生成多个扬声器的transmittance。注意和解码器模块负责从重叠的语音中产生多个transmittance。与PIT相比,SOT有两个优点:(1)对说话者的最大数量没有限制,以及(2)能够对不同说话者的输出之间的依赖关系进行建模。我们还提出了一个简单的技巧,允许SOT被执行在$O(S)$,其中$S$是在训练样本中的扬声器的数量,通过使用的组成源话语的开始时间。在LibriSpeech语料库上的实验结果表明,SOT模型在不同说话人数目的重叠语音的转录上明显优于基于PIT的模型。我们还表明,SOT模型可以准确地计算输入音频中的扬声器数量。

英文摘要:

This paper proposes serialized output training (SOT), a novel framework for multi-speaker overlapped speech recognition based on an attention-based encoder-decoder approach. Instead of having multiple output layers as with the permutation invariant training (PIT), SOT uses a model with only one output layer that generates the transcriptions of multiple speakers one after another. The attention and decoder modules take care of producing multiple transcriptions from overlapped speech. SOT has two advantages over PIT: (1) no limitation in the maximum number of speakers, and (2) an ability to model the dependencies among outputs for different speakers. We also propose a simple trick that allows SOT to be executed in $O(S)$, where $S$ is the number of the speakers in the training sample, by using the start times of the constituent source utterances. Experimental results on LibriSpeech corpus show that the SOT models can transcribe overlapped speech with variable numbers of speakers significantly better than PIT-based models. We also show that the SOT models can accurately count the number of speakers in the input audio.

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。

↑