一种基于线性预测与自相关函数法的语音基音周期检测新算法

阐述了一种新的基于语音线性预测模型和经典自相关函数法的基音周期检测算法。该算法的特点在于着重对被处理语音进行滤波预处理和平滑后处理。它克服了自相关法在对低信噪比的语音信号检测时常出现半倍频误差的缺点，同时又克服了对某些变化曲线较大的语音信号检测时随机错误较多的问题。实验表明，在噪声环境下．这种处理方法能够提高传统自相关基音检测法的可靠性和准确性。

A New Method for Speech Signals Pitch Detection Based on LPC and Autocorrelation

In this paper,a new pitch detection method based on classical autocorrelation function is proposed.This method's feature lies in its pre-procession and post-procession.It overcomes the disadvantage of the auto-correlation function-the half and double frequency error of ten happened in the low SNR environm ents.It also overcomes the stochastic error problem when the speech signals have a comparatively big change curve.The result of the experiment shows that in noisy environments these processions can im prove the reliability and accuracy of classical pitch extraction method.