基于LV-AMDF的自适应基音检测算法研究

根据语音信号的基音周期范围有限和周期相对稳定的特点，改进了可变长平均幅度差函数法(LV AMDF)，提出一种自适应幅度差法检测基音周期。它在语音非稳定段通过简单的谷值点评选机制，筛选当前谷值点以及历史谷值点，得到较精确的基音周期；在语音稳定段依据历史谷值点缩短语音段的比较范围，减少计算代价。还改进了浊音起止点检测算法，使浊音起止点的定位更精确。实验证明，该方法在不同的信噪比环境下有效地降低了半周期和倍周期点的发生率。

Adaptive speech pitch detection based on LV AMDF

According to the limitation and the stability of the speech pitch, the LV AMDF method was improved, and an adaptive speech pitch determination method was proposed. In the unstable speech, the exact speech pitch was acquired according to the trough in the current window and the window before by the simple trough select system; in the stable speech, the times of the speech pitch comparison was reduced, and the computing cost was decreased. The method for searching the beginning and ending point of the sonant was improved. Thus, more accurate points could be acquired. Simulations show that the pitch extracting results of the improved method reduces the rate of the half and double speech pitch points in different SNR environment.