摘要指挥训练系统在军事、航空、公安等领域都有广泛应用,而使用在船舰上的各类调度指令都需要一定的精确度,利用语音合成和识别技术可以使训练的效率提高。国内外在这一领域所做的研究有很多,从算法的研究到模型的建立,最后的框架趋向完善化。本课题对小样本基于Sphinx平台进行中文语音识别研究,在搭建的系统中根据船舰指挥指令规范,以及一些语音模型工具,设计一个针对船舰指挥的语音识别系统。运用Java编程等技术,通过测试得到成果证明,该系统的识别率较高,有效提高了船舰指挥系统的训练效率。19577
关键词 指挥系统 语音识别 语言模型 Java编程 识别率
毕业设计说明书(论文)外文摘要
Title Command Training System About Design & Realization of Mandarin Speech Recognition Technology
Abstract
Command Training System is used widely in military aeronautics management area and etc.And those instruction scheduling used in the naval vessel need to be accurate.Using Speech synthesis and recognition technology can increase the training efficiency. There is a lot of research in this field at home and abroad have done, from the establishment of algorithm to model building, at last the frame tend to perfection.This project use speech recognition at small sample on the basis of Sphinx platform. In setting up the system, according to the Command Training directive and some speech model tools design a naval vessel command speech recognition system, using Java programming technology.The experimental results indicate that the system has a high recognition rate, improving the training efficiency of the system to a great extent.
Keywords Training system Speech recognition Language model Java programming Recognition rate
目 次
1 绪论 1
1.1课题研究背景及意义1
2 系统搭建 3
2.1语音识别引擎Sphinx-4 3
2.2Sphinx-4系统架构 3
3 船舰指挥指令规范 5
3.1指令的语音特征分析 5
3.2 特殊语音识别5
4 声学模型的训练6
4.1建立声学模型6
4.2隐马尔科夫模型6
4.3声学模型的训练6
4.4中文训练的注意事项7
5 语音模型的训练9
5.1 N-Gram模型9
5.2 Cmulmtk工具9
6 设计语音识别系统10
6.1语音识别系统设计流程10
6.2训练环境10
6.3 训练步骤11
7 实验结果及分析15
7.1中文的字和词的问题15
7.2测试结果分析15
7.3试验界面预览16