语音信号准确性识别仿真研究
|
摘要
对语音信号的准确性进行有效识别,可提升语音发音准确性。但是语音信号中包含大量冗余度很高的随机信号,需要对随机信号进行过滤,降低降低信号的冗余度。但是传统方法利用小波变换对语音信号进行适当的尺度分解,得到各尺度的频谱并进行DCT变换不同语音信号的特征参数完成识别,但是无法排除随机信号的干扰,存在语音信号准确性识别误差大的问题。提出一种改进复子波的语音信号准确性识别方法。依据小波变换基本原理对含噪声的语音信号进行分解,获取信号不同分解层次的子带区间小波熵,并对信号不同层次中高频系数阈值进行计算,构建折中指数阈值函数,与计算结果相结合去除语法语音信号中高频系数所含的噪声,融合于复子波分析原理,计算出语言发音信号的基音周期及共振峰信息,利用最优的复高斯子波提取语音信号的幅度谱和相位谱,并完成对语音信号准确性识别。仿真结果表明,所提方法可以有效地降低信号冗余度,可以对语音信号的准确性有效识别。
This article proposes a recognition method for speech signal accuracy based on modified complex wavelet.Firstly,the method decomposes speech signals with noise according to wavelet transform fundamental.It acquires wavelet entropy of signal sub- band section in different decomposition orders and calculates threshold of high frequency coefficient.Then,it builds threshold function of compromise index to eliminate noise contained by the high frequency coefficients in speech signal of grammar integrated with then calculation results.It also workes out pitch period and formant information of signal of language pronunciation integrated with the complex wavelet theory.Finally,it uses optimal complex Gauss wavelet to extract magnitude spectra and phase spectra of signal and completes recognition on accuracy of speech signal.Simulation results show that the method can reduce signal redundancy effectively.It can recognize the signal accuracy effectively.
This article proposes a recognition method for speech signal accuracy based on modified complex wavelet.Firstly,the method decomposes speech signals with noise according to wavelet transform fundamental.It acquires wavelet entropy of signal sub- band section in different decomposition orders and calculates threshold of high frequency coefficient.Then,it builds threshold function of compromise index to eliminate noise contained by the high frequency coefficients in speech signal of grammar integrated with then calculation results.It also workes out pitch period and formant information of signal of language pronunciation integrated with the complex wavelet theory.Finally,it uses optimal complex Gauss wavelet to extract magnitude spectra and phase spectra of signal and completes recognition on accuracy of speech signal.Simulation results show that the method can reduce signal redundancy effectively.It can recognize the signal accuracy effectively.
引文
[1]杨鸿武,赵涛涛.一种基于加权Mel倒谱的语音信号共振峰提取算法[J].西北师范大学学报(自然科学版),2014,50(1):53-57.
[2]郑尚新,曹梦霞.语音信号中基频提取方法研究与综述[J].电脑与信息技术,2014,22(2):8-10.
[3]杨立东,谷宇,张明.语音信号特征选择优化提取仿真研究[J].计算机仿真,2016,33(2):409-412.
[4]孙颖,等.基于混沌特性的情感语音特征提取[J].天津大学学报(自然科学与工程技术版),2015,48(8):681-685.
[5]赵涛涛,杨鸿武.结合EMD和加权Mel倒谱的语音共振峰提取算法[J].计算机工程与应用,2015,51(9):207-212.
[6]鲜晓东,樊宇星.基于Fisher比的梅尔倒谱系数混合特征提取方法[J].计算机应用,2014,34(2):558-561.
[7]张明,刘祥楼,姜峥嵘.基于LPC的语音信号预测仿真分析[J].光学仪器,2015,37(1):71-74.
[8]夏晓玲.集成语音信号与表面肌电信号的情感自动识别模型[J].激光杂志,2015,36(4):131-134.
[9]米川,白俊贤.基于线性预测分析的语音信号合成[J].通信电源技术,2014,31(6):80-82.
[10]何昕,蒋豪,韩丹.管制指令特征参数提取研究[J].科学技术与工程,2015,15(20):89-94.
[2]郑尚新,曹梦霞.语音信号中基频提取方法研究与综述[J].电脑与信息技术,2014,22(2):8-10.
[3]杨立东,谷宇,张明.语音信号特征选择优化提取仿真研究[J].计算机仿真,2016,33(2):409-412.
[4]孙颖,等.基于混沌特性的情感语音特征提取[J].天津大学学报(自然科学与工程技术版),2015,48(8):681-685.
[5]赵涛涛,杨鸿武.结合EMD和加权Mel倒谱的语音共振峰提取算法[J].计算机工程与应用,2015,51(9):207-212.
[6]鲜晓东,樊宇星.基于Fisher比的梅尔倒谱系数混合特征提取方法[J].计算机应用,2014,34(2):558-561.
[7]张明,刘祥楼,姜峥嵘.基于LPC的语音信号预测仿真分析[J].光学仪器,2015,37(1):71-74.
[8]夏晓玲.集成语音信号与表面肌电信号的情感自动识别模型[J].激光杂志,2015,36(4):131-134.
[9]米川,白俊贤.基于线性预测分析的语音信号合成[J].通信电源技术,2014,31(6):80-82.
[10]何昕,蒋豪,韩丹.管制指令特征参数提取研究[J].科学技术与工程,2015,15(20):89-94.