【作者】 邓立新； 杨震； 郑宝玉；

【Author】 Deng LiXin Yang Zhen Zheng BaoYu (Department of Information Engineering of Nanjing University of Post and Telecommunications,P.O.Box 165,210003)

【机构】 南京邮电学院信息工程系；

【摘要】 本文提出了一种判别语音静、清、浊三态S/U/V的新方法——语音波形图象模板匹配法。它的基本原理是首先利用语音信号的波形图,来导出静、清、浊三态的标准矢量模板,然后将输入信号波形转换成矢量图,与标准模板进行比较。此法不仅简单、高效,而且对于过渡帧的判别精度高于其它一些判别法,并具有一定的抗突发噪音性能。更多还原

【Abstract】 In this paper,we propose a new method of Voiced/Unvoiced/Silence classifications of speech signal-Image Model Matching Method of Speech Waveform.It’s basic principle is that firstly the standard vector model of each class(U/V/S) is deduced by exploiting corresponding wave image of speech signal,then we transfer the input waveform of one frame signal into its corresponding vector and compare it with the standard vector model to obtain the classifying result.In addition to the simplicity and high efficiency,this new method has higher precision in classifying transitional frame than some other classification methods.Also,It has performance of resisting sudden noise.更多还原