基于深度神经网络的蒙古语声学模型建模研究

PDF下载HTML

Title:

Mongolian acoustic modeling based on deep neural network

作者:

马志强, 李图雅, 杨双涛, 张力

内蒙古工业大学 数据科学与应用学院, 内蒙古 呼和浩特 010080

Author(s):

MA Zhiqiang, LI Tuya, YANG Shuangtao, ZHANG Li

School of Data Science & Application, Inner Mongolia University of Technology, Hohhot 010080, China

关键词:

语音识别; 声学模型; GMM-HMM; DNN-HMM; 监督学习; 预训练; 过拟合; dropout

Keywords:

speech recognition; acoustic model; GMM-HMM; DNN-HMM; supervised learning; pre-training; over-fitting; dropout

分类号:

TP391

DOI:

10.11992/tis.201710029

摘要:

针对高斯混合模型在蒙古语语音识别声学建模中不能充分描述蒙古语声学特征之间相关性和独立性假设的问题，开展了使用深度神经网络模型进行蒙古语声学模型建模的研究。以深度神经网络为基础，将分类与语音特征内在结构的学习紧密结合进行蒙古语声学特征的提取，构建了DNN-HMM蒙古语声学模型，结合无监督预训练与监督训练调优过程设计了训练算法，在DNN-HMM蒙古语声学模型训练中加入dropout技术避免过拟合现象。最后，在小规模语料库和Kaldi实验平台下，对GMM-HMM和DNN-HMM蒙古语声学模型进行了对比实验。实验结果表明，DNN-HMM蒙古语声学模型的词识别错误率降低了7.5%，句识别错误率降低了13.63%；同时，训练时加入dropout技术可以有效避免DNN-HMM蒙古语声学模型的过拟合现象。

Abstract:

Considering the difficulty of using the Gaussian mixture model (GMM) to adequately describe the correlation and independence hypothesis of the Mongolian acoustic features in the acoustic modeling of Mongolian speech recognition, this study investigates an acoustic model based on deep neural network (DNN). Firstly, using DNN, the internal structure of phonetic features were classified and learned to extract the Mongolian acoustic features, and a DNN-HMM Mongolian acoustic model was constructed. Secondly, a training algorithm was designed by combining unsupervised pre-training and supervised training tuning. In addition, dropout technology was added into the DNN-HMM Mongolian acoustic model training to avoid the over-fitting phenomenon. Finally, a comparative experiment was conducted for the GMM-HMM and DNN-HMM Mongolian acoustic models on basis of the small-scale corpus and Kaldi experimental platform. Experimental results show that the word recognition error rate of DNN-HMM Mongolian model was reduced by 7.5% and sentence recognition error rate was reduced by 13.63%. In addition, the over-fitting of DNN-HMM Mongolian acoustic model can be effectively avoided by adopting the dropout technique during training.