针对传统的基音估计算法不能适用于过渡音的问题,提出中心偏离函数及基于中心偏离函数的基音周期检测算法。该算法主要通过划分语音子带来搜索最小中心偏移量,较准确地确定过渡音的基音周期。介绍了该算法的设计思想和实现过程,并给出代表性的实验结果,将其结果与MBE算法所检测到的基音周期进行比较。研究结果表明:采用所提出的中心偏离函数和时域削波方法相结合来搜索基音周期,能够有效地克服传统方法的不足,降低误判率,提高基音周期的估计精度。
A central bias function and an arithmetic of pitch detection based on central bias function were proposed for the transition sound, which cannot be detected by the traditional arithmetic of pitch calculation. The arithmetic searches for the least central offset through partition speech sub-band and confirms pitch period of the transition sound. The implementation process and the designing idea of the arithmetic were described in detail, and the representative experiment resuits were compared with that obtained by MBE arithmetic. The results show that the method using bias function and time-domain clipping to search pitch period can overcome shortcomings of the traditional, reduce false decision and raise the estimation precision.