基于TD-PSOLA普通话韵律参数修改的实现  

Realization of Chinese Speech Prosodic Parameter Modification Based on TD-PSOLA

作　　者：张后旗[1] 傅卫卫[1] 方志刚[2] 

机构地区：[1]浙江财经学院信息分院,浙江杭州310018 [2]浙江大学城市学院,浙江杭州310015

出　　处：《计算机仿真》2009年第7期329-332,共4页Computer Simulation

摘　　要：在普通话文语转换系统中,如何精确控制韵律参数修改是关系到合成语音质量的关键。根据汉语普通话的韵律特征,采用TD-PSOLA算法,精确实现了的韵律参数的修改。用分时段直线韵律参数模型,动态生成基音轮廓,计算出合成基音标记,一次性实现基频尺度修改和时间尺度的修改。根据自然语音与模拟合成出来的语音韵律参数比较分析表明,能有效地控制韵律参数,达到预期的参数合成目标,并且具有较高质量。也可以进行进行参数合成运算量少,控制参数精确,合成语音清晰自然。In Chinese Speach Text - To - Speech system, how to precisely control the prosodic parameter modification is the key to the quality of synthesis speech. Based on Chinese Speech prosodic feature, an effective Chinese speech prosodic parameter modification method using TD - PSOLA algorithm is precisely realized. This paper puts forward a time - stage line fuction prosodic parameter model, produces synthesis pitch contour temporarily, and calculates synthesis pitch masks with time - scale and pitch - scale modified simultaneously. Comparing the synthetic speech with the autual speech, it＇ s found that this method can control prosodic parameters effectively, and achive disired projective, and the quality of synthtic speech is high. Using this method, it can conclude that there are sevreal advantages sucn as less amount of operation, precise parameter modification and high intelligibility and naturalness.

关 键 词：时域基音同步叠加算法 基音轮廓 韵律参数 时间弯曲函数 语音合成 

分 类 号：TN912.33[电子电信—通信与信息系统]