摘要
为了提高流媒体环境下英语口语发音自动校对能力,提出一种基于群延迟谱特征提取和智能语音识别的英语口语发音自动校对方法,构建流媒体环境下英语口语发音信号模型,采用声传感器进行流媒体环境下英语口语发音信号采集,对采集的语音信号采用群延迟估计方法进行时域相关性补偿和滤波检测,提取流媒体环境下英语口语发音信号的多参数统计特征量,采用连续数字语音识别方法进行英语口语发音的谱特征提取,分析提取的谱特征量的频谱差异性,采用语音结构化模型进行发音准确性校对和自适应反馈,提高了流媒体环境下英语口语发音自动校正能力。仿真结果表明,采用该方法能有效实现流媒体环境下英语口语发音自动校对,对语音信号的检测和参数估计性能较好。
In order to improve the automatic proofreading ability of oral English pronunciation in streaming media environment,a new method based on group delay spectrum feature extraction and intelligent speech recognition is proposed to automatically proofread spoken English pronunciation,and a speech pronunciation signal model in streaming media environment is constructed.The acoustic sensor is used to collect the spoken English pronunciation signals in streaming media environment,and the group delay estimation method is used to compensate the correlation in time domain and filter detection.The multiparameter statistical features of spoken English pronunciation signals in streaming media environment are extracted.The spectral features of spoken English pronunciation are extracted by continuous digital speech recognition method,and the difference spectrum of the extracted spectral features is analyzed.Pronunciation veracity proofreading and adaptive feedback based on structured speech model can improve the ability of automatic pronunciation correction in streaming media environment.The simulation results show that this method can effectively realize the automatic pronunciation of spoken English in streaming media environment,and the performance of speech signal detection and parameter estimation is better.
作者
牛腊婷
NIU Lating(Shaanxi Energy Institute,Xianyang Shaanxi 712000,China)
出处
《自动化与仪器仪表》
2020年第7期155-158,共4页
Automation & Instrumentation
基金
陕西省自然科学基础研究计划资助项目(No.2014JM8346)。
关键词
流媒体环境
英语口语
发音
自动校对
谱特征分析
streaming media environment
spoken English
pronunciation
automatic proofreading
spectral feature analysis
