基于改进算法的基音检测及其应用

Application of Improved Algorithm based on Pitch Detection

语音信号的基音周期在语音压缩编码、语音分析与合成、语音识别和说话者确认等方面是很重要的参数。到目前为止,还没有一种方法能在不同的话者、不同的环境下都达到最佳的效果。根据汉语语音信号的特点,本文首先对语音进行滤波处理、中心削波处理来滤除语音中高次谐波成分及高频噪声,接着利用修正自相关函数提取语音基音周期,最后对所得的结果进行平滑处理。实验结果显示改进算法提高了基音提取的可靠性,最后本文在改进算法的基础上进行了声调识别。

BSTRACT Pitch period is a crucial speech parameter in many applications such as speech compress coding,analysis and synthesis,speech segmentation and automatic isolated word recognition etc.Until now,there is no method that can detect the pitch period perfect in any conditions.In the light of the properties of Chinese speech,this paper presents a method with short-time auto-correlation to detect the pitch,then the paper gets the tone based on the pitch.