The Unvoiced/Voiced Segment Method Based on Granular Computing

张起贵

您当前的位置：

首页 >

文章列表页 >

The Unvoiced/Voiced Segment Method Based on Granular Computing

更新时间：2026-09-14

- The Unvoiced/Voiced Segment Method Based on Granular Computing
- 基于粒计算的清浊音分段算法
- Vol. 30, Issue 2, Pages: 184-188(2009)
- 作者机构：
  
  太原理工大学信息工程学院,山西,太原,030024
- 作者简介：
- 基金信息：
- DOI：
  CLC：
- Published：2009
- 稿件说明：
移动端阅览
张起贵. The Unvoiced/Voiced Segment Method Based on Granular Computing[J]. 2009, 30(2): 184-188.
DOI：

张起贵. The Unvoiced/Voiced Segment Method Based on Granular Computing[J]. 2009, 30(2): 184-188. DOI：

摘要

评价一个参数对于划分语音信号成清浊音段的有效性是十分困难的问题.为解决此问题

提出了一种基于粒计算的语音信号清浊音检测算法.该算法利用粒的理论对清浊音检测中用到的8种语音信号的特征参数进行分析

构造决策表

计算得出各参数对清浊音判决的重要度.最终选用参数组合MFCC(Mel-Frequence Ceptral Coefficients)、小波互相关、小波系数和短时自相关进行语音实时分段

得到多参数结合的求解最优路径.进一步提取决策规则并进行多参数清浊音分段

试验表明获得了好的效果:对于纯净语音分段算法的准确率达到平均93.86%

对于带噪语音准确率达到平均90.27%.此外算法的处理帧长仅为20样点

意味着有很好的实时性.

Abstract

It is a difficult problem how to evaluate a parameter's effectiveness for dividing a speech signal into unvoiced or voiced segment.An application of granular computing to this topic is presented.By analyzing eight parameters of the speech signal

a decision table is constructed depending on the granulation theory

which provides a method for evaluating the significance of parameters to judge the unvoiced/voiced decision.And then the optimum scheme for real-time speech signal segment and making decision rules is obtained by choosing multi-parameter set MFCC

wavelet cross-correlation

wavelet coefficient and short-time selfcorrelation.The experimental results indicate that the accuracy of segmentation algorithm is average 93.86% for pure speech and 90.27% for speech mixed with noise.In addition

the update frame size is 20 samples in this algorithm

which denotes its good real-time performance.

关键词

Keywords

references

Rabiner L R,Sambur M R.An Algorithm for determining the endpoint of isolated utterance[J].Bell System TechJ,1975,54(2):297-315.

周志杰,胡光锐,林晓彤.采用Fisher分类法实现清浊音判决[J].上海交通大学学报,1997,31(5):30-32.Zhou zhijie,Hu guangrui,Lin xiaotong.Realizing unvoiced/voiced speech decision by Fisher classification[J].Journal of Shanghai Jiaotong University,1997,31(5):30-32.(in Chinese)

于俊凤,曹俊兴.基于组合参数的清浊音判决方法[J].太原理工大学学报,2004,35(4):467-469.Yu junfeng,Cao junxing.Unvoiced/voiced decision method based on combination parameter[J].Journal of TaiyuanUniversity of Technology,2004,35(4):467-469.(in Chinese)

张永,丁洪昌.连续属性离散化的MaxDiff方法[J].计算机工程与应用,2007,43(19):80-82.Zhang yong,Ding hongchang.The MaxDiff method of continuous attribute discretization[J].Computer Engineeringand Application,2007,43(19):80-82.(in Chinese)

张文修,吴伟志,梁吉业,等.粗糙集理论与方法[M].北京:科学出版社,2000.

李道国.信息粒计算理论、模型与应用研究[M].太原:山西科学技术出版社,2006.

陈玉明.基于信息粒与粒计算理论的数据约简研究[D].南昌:南昌大学,2005.

郝静.基于粒计算的语音实时分段算法[D].太原:太原理工大学,2008.

Views

108

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Configuration Design and Research of a Hybrid Quadruped Perilla Sowing Robot

High⁃Precision Photoeletric Coordinate Positioning Technology Based on Transformer⁃FPGA Collaborative Feature Modeling

Study on the Ultrasonic Adsorption of Cr（Ⅵ） in Water by Tea Biochar Magnetic Materials

Research Progress on Single⁃Photon Detection for Quantum Information About Mechanisms，Devices and Circuit Design

A Meta⁃Learning and VAE⁃GAN Based META⁃VAEGAN⁃DNN Method for Network Intrusion Detection

Related Author

No data

Related Institution

No data

⁰