智能仪器嵌入式声纹识别技术方法  

Intelligent Instrument Embedded Voice Recognition Technology

作　　者：于哲舟[1] 杨佳东[1] 周春光[1] 王纲巧[1] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012

出　　处：《仪器仪表学报》2004年第z3期447-450,共4页Chinese Journal of Scientific Instrument

摘　　要：在仪器上嵌入语音识别技术对提高仪器的智能性、方便性和玩具的趣味性具有重要意义和作用。利用嵌入式技术与仪器仪表的制造技术相结合的方法广泛地应用于各个领域,德国开发的汽车车载声控系统、日本研制的能识别主人声音的智能玩具等。安防仪器中嵌入语音识别技术的作用更为突出,是解决身份被盗、证件丢失及假冒等造成安全问题的很有效方法。嵌入式声纹识别门禁系统是在对语音识别技术进行了大量研究,对这个领域主要技术、工作流程及各部分采用的方法进行细致分析的基础上提出的。该系统是对传统隐马尔科夫模型Baum - Welch训练算法的改进,同时提出“多方法多门限关联识别”技术,部分解决了说话人识别中“阈值设置”这一难点,提高了正识率与系统效率。该系统嵌入到门锁中,体积小、智能程度高。试用证明了此方法对系统改进的有效性,为开发智能化仪器提供了一个较好技术和方法。It is important to embed the voice recognition technique on the instrument for improving the intelligence and convenience of the instrument and the interest of the toys.The combination of the embedded technique with various apparatus and instrument is applied to various fields widely.the voice-driven system on car which is developed by Germany,the intelligent toys which are developed by Japanese and so on.It is more important that the voice recognition technique embedded in security instruments,and this method is an effective solution to solve the stolen identification card,the lost certificate and other security problems.This paper proposed the embed voice door manager system based on the research and discussion of voice recognition technique and the analysis of the main technique,work flow and the method of each part.This system modified the traditional Baum-Welch training algorithm of hide Markov model,and proposed various method and various threshold connected recognition technique.Those techniques partly solve the key technique of the threshold-setting,improve the correctly recognition rate and the efficiency of the system.This system is embedded in the door lock,and it has small size and high intelligence.The trial application proves that this method is valid to improve the system,and provides a good technique and method for developing intelligent instruments.

关 键 词：智能仪器 语音识别 嵌入式 隐马尔科夫模型 

分 类 号：TH7-55[机械工程—仪器科学与技术]