语音信号时频特征显示系统的设计和仿真  

Design and simulation of time-frequency features display system for speech signal

作　　者：王光艳[1,2] 赵晓群[3] 王霞[1] 

机构地区：[1]河北工业大学信息工程学院,天津300130 [2]天津商业大学信息工程学院,天津300134 [3]同济大学电子与信息工程学院,上海200092

出　　处：《计算机工程与应用》2010年第29期73-75,97,共4页Computer Engineering and Applications

基　　金：天津市高校科技发展基金(No.20080710)

摘　　要：语音信号处理算法众多,但用于语音处理算法验证和开发的可视化研究平台极少。基于MATLAB GUI技术,完成语音信号典型时频特征参数提取和显示系统仿真平台的设计。可实现多种格式音频文件的载入和播放、波形和频谱显示、以及线性预测倒谱系数和美尔倒谱系数的计算、存储和显示等功能。系统界面友好、操作方便,可实现参数的交互输入并控制显示结果。仿真结果验证了相关时频特征参数提取算法的正确性,提高了对算法或数据处理效果的直观认识。Although the algorithms for speech processing are plentiful,there are few visualization software platforms which can be used to testify or design some specific algorithms.In this paper, a system platform used for extracting and indicating the time-spectral features of speech signal is designed under the environment of MATLAB GUI.On this platform, the user can load and play the speech signal in different audio formats.Especially,Linear Prediction Cepstrum Coefficient（LPCC） and Mel-Frequency Cepstrum Coefficient（MFCC）,which are the two main feature parameters for speech recognition system,can be calculated,saved and displayed on the same panel.The system interface is friendly,as well as it is convenient and interactive for operating.The simulation results demonstrate the accuracy of the background algorithms for time-frequency features extracting.The designed system will provide important and intuitive auxiliary effect on verifying the algorithms and data processing efficiency for the research fields related to speech signal processing.

关 键 词：时频特征 频谱 线性预测倒谱系数 美尔倒谱系数 

分 类 号：TN912.34[电子电信—通信与信息系统]