摘要: 针对现有语音基音检测算法抗噪能力低的问题,利用多尺度连续小波对语音信号进行预处理。基于短时平均幅度差函数和短时自相关函数,提出一种语音基音周期检测方法,在确保运算量较低的前提下,获得准确结果。使用平滑算法消除倍频点、半频点和随机点产生的误差,实现基音曲线平滑。
关键词:
连续小,
尺度分析,
音周期检测,
相关函数
Abstract: Current pitch period detection algorithms can not work well under high noise environment. Aiming at this problem, this paper utilizes multi-scale continuous wavelet to pre-filtering the signal of voice. Based on the short-time autocorrelation function and the average magnitude difference function, a method for voice pitch period detection is proposed. This method needs little computation expense and can obtain the right pitch period. Smoothing algorithm is exploited to remove the errors of multiple frequency point, half frequency point and random point, and realize pitch curve smoothing.
Key words:
continuous wavelet,
multi-scale analysis,
tch period detection,
utocorrelation function
中图分类号:
郑继明, 王劲松. 语音基音周期检测方法[J]. 计算机工程, 2010, 36(10): 273-275.
ZHENG Ji-Meng, WANG Jin-Song. Method for Voice Pitch Period Detection[J]. Computer Engineering, 2010, 36(10): 273-275.