一种语音分段新算法研究  

Study on New Algorithm of Segment Vocoder

作　　者：王军强[1] 安云峰[2] 朱章华[3] 

机构地区：[1]西北工业大学第365研究所,陕西西安710072 [2]西安通信学院通信指挥系,陕西西安710106 [3]重庆通信学院机动作战通信系,重庆400035

出　　处：《测控技术》2009年第2期28-30,共3页Measurement & Control Technology

摘　　要：分段声码器是一种高效的低速语音编码方法,针对分段声码器实现的关键技术语音分段方法进行了研究,提出了一种分段新算法。该算法通过检测临界带特征矢量参数的距离变化确定初始分段边界,结合动态规划方法提高分段的准确性。仿真结果表明:新算法实用有效,相对于以前分段声码器中应用的STM方法和最大似然方法,语音分段性能得到提高。Segment vocoder is an efficient very low bit rate speech coding algorithm. Speech segment method, one of the key technology of segment vocoder, is studied and a new algorithm is presented. The initial segment is determined by detecting the distance changes of the critical feature parameter vector. With the dynamic pro- gramming method, the segmental accuracy is improved. Simulation shows the new algorithm is efficient and performs better than the STM method used in the segment vocoder before and the ML method.

关 键 词：分段声码器 语音分段 距离测度 动态规划 

分 类 号：TN912[电子电信—通信与信息系统]