声纹识别在虚拟仪器平台的实现  

On implementation of voiceprint recognition in virtual instrument platform

作　　者：王会清[1,2] 张涛[2] 周帆[2] 

机构地区：[1]武汉工程大学智能机器人湖北省重点实验室,湖北武汉430074 [2]武汉工程大学计算机科学与工程学院,湖北武汉430074

出　　处：《武汉工程大学学报》2012年第12期58-61,共4页Journal of Wuhan Institute of Technology

摘　　要：采用LabVIEW软件和声卡构建一个声纹识别系统.通过对音频信号的采集及处理,分析说话人的声音,提取出语音的频率时间序列作为特征值,建立模型库.识别时,由系统对现场输入的语音采样、提取特征值,并与注册样本模型进行匹配,根据其结果作出是否一致的判断.系统测试结果表明,用提取文本语音频率时间序列进行身份认证是一种能满足实际要求的可行方法.该方法可以区分不同的说话人,能有效地克服环境噪声带来的影响,并且在做说话人确认测试时达到较高的识别率.在采用麦克进行语音录入与匹配时,学习、识别时间短,声纹确认准确,声纹确认的漏识率小于1%.具有无区域限制,简单方便,成本低等特点,可以为声纹识别系统的应用提供一个安全可靠的技术支持.A voiceprint recognition system was built using LabVIEW and sound card. Through the acquisition and processing of the audio signal, including analyzing the speaker＇s voice and extracting voice frequency-time sequence as the characteristic value, the model library was created. During the authentication, the on-site input voice was sampled by the system and extracted the characteristic value, and matched with the registered sample model, thus a consistency judgment was made based on the result. System test results show that the identity verification by extracting text-to-speech frequency-time series is a feasible method to meet the actual requirements. The method can distinguish between different speakers, effectively overcome the impact of environmental noise, and achieve higher recognition rate in the speaker verification test. Using the microphone for voice input and matching, learning and recognition time is short, confirmation by the voiceprint is accurate, and the ratio of missing recognition is less than 1%. With the advantages of no regional restrictions, simplicity and convenience, and low cost, the system can provide a safe and reliable technical support for the application of the voiceprint recognition system.

关 键 词：虚拟仪器 声纹识别 特征提取 说话人确认 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]