噪声环境下的汉语声调识别  

Chinese Tone Recognition under Noise Conditions

作　　者：顾明亮[1] 夏玉果[2] 王劲松[1] 

机构地区：[1]徐州师范大学物理系,江苏徐州221116 [2]江苏省语言科学与神经认知工程重点实验室,江苏徐州221116

出　　处：《计算机技术与发展》2007年第8期70-72,76,共4页Computer Technology and Development

基　　金：江苏省"十五"社科基金项目(K3-013);江苏省高校自然科学基金(99KJB510002)

摘　　要：利用语音信号与噪声信号具有不同相关特性的特点,提出了一种新的加权自相关基频检测算法,该方法可以提高噪声环境下基音检测的准确性。在分类器设计方面,通过引入支持矢量机,进一步提高低信噪比下的汉语声调识别率。实验结果表明,新方法对提高噪声环境下的声调识别效果是十分有效的。According to the fact that speech and noise have different correlation character, presents a novel pitch extraction algorithm based on the weigthed correlation. This approach can raise the correction of pitch extraction for noise speech. On the design of classifier, by introducing support vector machine,Chinese tone recognition rate is improved under low signal noise rate conditions. The experimental resuits show that these improvement methods is very useful .

关 键 词：基音检测 支撑矢量机 汉语声调识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]