摘要
基音周期搜索的准确性将直接影响到语音编码器的编码质量和效率。本文根据AMR-WB+标准中基音周期搜索算法会发生检测基音倍频和半频错误,提出了开环基音搜索算法。该算法以自相关函数为基础,利用基音周期的平滑性,引入基音周期全局参考作为基音周期判断的辅助条件,有效解决了基音周期加倍的问题并在基音周期预测中体现基音周期的平滑性,实验结果表明本文算法性能优于AMR-WB+中的算法性能。该算法已应用到AVS-P10移动音频编解码框架中。
Pitch detection is ciucial importance for speech coding, and also is a critical factors to improve the performance of speech codec. The pitch detection algorithm of AMR - WB + has an problem of detecting a half or double frequency pitch. For that, the paper provides an algorithm of open - loop pitch detection. Based upon the smooth property of pitch, the provided approach calculates the correlation function, and introduces the global reference to contribute for estimation of pitch. The proposed approach solves the trouble of double pitch detection. The experiments shows the algorithm is novel and superior with AMR - WB + , Which has been applied to AVS - P10 for China.
出处
《智能计算机与应用》
2014年第1期75-77,82,共4页
Intelligent Computer and Applications
基金
深圳市生物
互联网
新能源
新材料产业发展专项资金基础研究计划(JC201104220203A)