汉语声调识别中的基音平滑新方法  

An Approach of Fundamental Frequencies Smoothing for Chinese Tone Recognition

作　　者：朱小燕[1] 王昱[1] 刘俊[1] 

机构地区：[1]智能技术与系统国家重点实验室清华大学计算机系,北京100084

出　　处：《中文信息学报》2001年第2期45-50,共6页Journal of Chinese Information Processing

基　　金：国家自然科学基金 !(6 9982 0 0 5 ) ;国家重点基础研究发展规划项目 !(G1980 30 5 0 70 3)

摘　　要：汉语普通话是一种带声调的语言。声调可以用基音的轮廓信息进行描述。传统基音的平滑方法 :线性平滑、中值平滑和一般的线性插值方法都不能很好地处理连续的基音频率有随机错误点的情况。本文提出了一种通过搜索来得到更精确的基音轮廓的新的基音平滑方法。这种方法具有简单可靠 ,快速高效的特点。实验表明这种方法比传统的方法识别错误率降低约 40 %Mandarin is a tonal language.The tones are recognized by using the pitch contour information which can be expressed by fundamental frequencies.The classic approaches for fundamental frequencies smoothing,such as linear smoothing,median smoothing and linear interpolation,can not work well in the case of that fundamental frequency is not detected correctly and several continulus frames.In this paper,a new smoothin approach is presented,in which a searching method is used to get a preferable accurate pitch contour.This approach is characterized by its simple,reliable and fast performance.Experimental results show that the new approach can decrease the recognition error rate by 40%.

关 键 词：基音检测 声调识别 语音识别 基音频率 基音轮廓 中值平滑 随机错误点 

分 类 号：TN912.34[电子电信—通信与信息系统]