基于深层神经网络的英语口语发音错误捕捉方法研究  

A deep Neural Network-based Method for Capturing Pronunciation Errors in Spoken English

作　　者：窦旭霞[1] Dou Xuxia(Yantai Vocational College,Yantai,Shangdong 264670,China)

机构地区：[1]烟台职业学院,山东烟台264670

出　　处：《黑龙江工业学院学报（综合版）》2020年第8期124-128,共5页Journal of Heilongjiang University of Technology(Comprehensive Edition)

基　　金：2019年烟台职业学院教学改革立项项目“‘互联网+’背景下创新高职英语课堂协作教学的实践研究”。

摘　　要：为了提高英语口语发音错误捕捉能力,提出基于深层神经网络的英语口语发音错误捕捉方法。构建英语口语发音信号检测模型,采用多传感融合跟踪识别方法进行语音信号采集,结合时频特征分解方法进行发音错误信息的特征提取,建立发音错误信号的统计特征分析模型,采用深层神经网络分类器进行发音错误信号的特征筛选和分类识别,实现英语口语发音错误捕捉。仿真结果表明,采用该方法进行英语口语发音错误捕捉的准确性较高,实用性强。In order to improve the ability to catch errors in spoken English,a method based on deep neural network is proposed.A model for detecting oral English pronunciation signals was established.The multi-sensor fusion tracking recognition method was used to collect the speech signals,and the time-frequency feature decomposition method was used to extract the pronunciation error information features.A statistical feature analysis model of pronunciation error signals is established,and a deep neural network classifier is used to screen and classify pronunciation error signals,so as to catch errors in spoken English.The simulation results show that this method is more accurate and practical in capturing errors in spoken English.

关 键 词：深层神经网络 英语口语 发音 错误 捕捉 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] H319[自动化与计算机技术—计算机科学与技术]