【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。更多还原

【Abstract】 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 results show that these improvement methods is very useful.更多还原