英语语音优化识别建模仿真分析  

Modeling and simulation of English speech optimization recognition

作　　者：米婧[1] MI Jing(Xianyang Vocational Technical College,Xianyang 712000,Shaanxi Province,China)

机构地区：[1]咸阳职业技术学院

出　　处：《信息技术》2019年第6期91-95,共5页Information Technology

摘　　要：随着中国经济高速发展以及全球一体化的进程,英语成为了人们日常交流必不可少的工具,然而对于初学者来说,能够通过语音识别技术将语音信号转化成文本的格式,更有利于快速掌握英语。而且语音识别技术经过多年的发展依然具有巨大的挖掘潜力,面对移动互联网的快速发展,通过对实时通信工具的大数据量的需求爆发,英语语音识别的实时性和系统稳定性越来越受到关注,文中分析了常用的传统语音识别技术,例如动态时间规整、神经网络模型和隐马尔可夫模型等,运用隐马尔可夫模型对语音信号进行处理和识别,提取出特征参数,与经过训练的模型体系进行匹配,找出最优的识别序列。然后在PC平台上,利用MATLAB建模仿真,基本实现了英语语音短句的识别,对于后续的硬件产品实现打下了良好的基础,具有积极的现实意义。With the rapid development of China’s economy and the process of global integration,English has become an indispensable tool for people’s daily communication.However,for beginners,voice signals can be converted into text formats through speech recognition technology,which is more conducive to the rapid mastery of English.And speech recognition technology still has tremendous potential after many years of development.Faced with the rapid development of mobile Internet,the real-time and system stability of English speech recognition have been paid more and more attention through the explosion of large data demand for real-time communication tools.This paper analyses the commonly used traditional speech recognition technologies,such as dynamic time regulation,neural network model and hidden Markov model,etc.l,and uses the hidden Markov model to process and recognize the speech signal,extract the characteristic parameters,match with the trained model system,and find out the optimal recognition sequence.Then on PC platform,using MATLAB modeling and simulation,the recognition of English speech short sentences is basically realized,which lays a good foundation for the realization of the hardware products behind,and has positive practical significance.

关 键 词：语音识别 发音质量评价 隐马尔可夫模型 特征参数提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]