吉林大学学报(理学版)

• 计算机科学 • 上一篇    下一篇

基于归一化算法的噪音鲁棒性连续语音识别

刘妍秀1, 孙一鸣2, 杨华民2   

  1. 1. 长春大学 教务处, 长春 130022; 2. 长春理工大学 计算机科学技术学院, 长春 130022
  • 收稿日期:2014-12-18 出版日期:2015-05-26 发布日期:2015-05-21
  • 通讯作者: 刘妍秀 E-mail:klxx123456@163.com

Noise Robust Continuous Speech Recognition Based on Normalization

LIU Yanxiu1, SUN Yiming2, YANG Huamin2   

  1. 1. Office of Academic Affairs, Changchun University, Changchun 130022, China; 2. College ofComputer Science and Technology, Changchun University
     of Science and Technology, Changchun 130022, China
  • Received:2014-12-18 Online:2015-05-26 Published:2015-05-21
  • Contact: LIU Yanxiu E-mail:klxx123456@163.com

摘要:

针对归一化方法在连续语音特征曲线调整时存在的问题, 提出一种优化解决方案, 解决了噪声的不稳定性及不可预测性对语音特征的影响. 结果表明, 基于该优化方法建立的鲁棒性连续语音识别模型可实现在实验室干净环境和现实噪音环境下同时得到较好的识别结果.

关键词: 归一化, 噪音鲁棒性, 连续语音识别

Abstract:

Analyzing the impact of normalization method applied in isolated word speech dominant and noise characteristics to discover the continuous speech characteristic curve adjustment problems. The authors raised optimized solutions to further solve the problem of instability and unpredictability of the noise characteristics for voice effects. Robust continuous speech recognition model by normalization method in this paper can achieve a clean environment in the laboratory and real noise environment so as to get the best recognition results.

Key words: normalization, noiserobust, continuous speech recognition

中图分类号: 

  • TP319