基于数学形态滤波的语音信号基音特征提取

Pitch extraction of speech signal based on mathematical morphological filter

摘要: 数学形态滤波是一种关于信号形状处理的非线性变换,它能简化信号、消除较小分量而保留信号的基本形状特征.本文基于数学形态滤波方法提出了两个分别在时域和频域提取语音信号基音周期的方案,在频域提取基音周期的同时还能提取出语音信号的谱包络。它们具有简单、直观和计算效率高等特点。由于数学形态滤波运算是并行的、局部的,新方案适于并行化处理和易于硬件化实现。实验结果表明,选择合理的数学形态滤波参数以及线性预测编码参数,能获得准确的语音信号基音特征。

Abstract: Mathematical morphological filtering is a nonlinear transform to operate on shape of the signal,which can simplify the signal and eliminate some small components of the signalwithout corrupting shape of the signal.In this paper,two schemes extracting pitch of speechsignal using mathematical morphological filtering are proposed,which are quite simple,compu-tationally very efficient,and apt to be implemented by hardware.Experimental results show that more precise pitch,as compared with conventional methods,will be obtained if appropriate parameters are selected.