基于MATLAB语谱图的声乐研究  

Research on Vocal Music Based on MATLAB Spectrogram

作　　者：白燕燕[1] 胡晓霞[1] BAI Yanyan;HU Xiaoxia(Xi'an University of Technological Information ,Xi'an 710200,China)

机构地区：[1]西安工业大学北方信息工程学院

出　　处：《软件工程》2019年第9期1-4,共4页Software Engineering

基　　金：西安工业大学北方信息工程学院2019年度院长科研基金资助项目(项目编号19YZ01)

摘　　要：语谱图是一种在语音分析以及语音合成中具有重要实用价值的时频图,能反映出语音信号动态频谱特征,被认定是语音信号的可视语言。横坐标代表时间大小,纵坐标代表频率大小,时间和频率所对应的像素点的值表示能量值的大小。本文对一小段语音信号和音乐信号进行处理,最终以语谱图的形式展现出来,并通过对语谱图的分析来判断语音信号和音乐信号,检测语音是否开始和结束,效果良好。这种方法容易区分语音信号中的有用声段和噪声声段。实验证明,纯语音信号频域能量和过零率变化要大于音乐信号。音乐信号较语音信号能量谱稳定。Spectrogram is a kind of time-frequency graph with important practical value in speech analysis and speech synthesis.It can reflect the dynamic spectrum characteristics of speech signals and is regarded as the visual language of speech signals.The horizontal coordinate represents time,the vertical coordinate represents frequency,and the value of pixels corresponding to time and frequency represents the magnitude of energy value.This paper deals with a small segment of speech signal and music signal,which are finally presented in the form of spectrogram.By analyzing the spectrogram,we can judge the speech signal and music signal,and detect whether the speech starts or ends.This method is easy to distinguish the useful sound segments from the noise segments in speech signals.It is proved that the change of frequency domain energy and zero crossing rate of pure speech signal are greater than that of music signal.The energy spectrum of music signal is more stable than that of speech signal.

关 键 词：语谱图 语音分析 语音合成 短时能量 过零率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]