首页> 外文期刊>WSEAS Transactions on Signal Processing >Optimization of hidden Markov model with Gaussian mixture densities for Arabic speech recognition
【24h】

Optimization of hidden Markov model with Gaussian mixture densities for Arabic speech recognition

机译:利用高斯混合密度的隐马尔可夫模型优化阿拉伯语语音识别

获取原文
获取原文并翻译 | 示例
           

摘要

Speech recognition applications are becoming more and more useful nowadays. In automatic speech recognition (ASR) systems, hidden Markov models (HMMs) have been widely used for modeling the temporal speech signal. Iterative algorithms such as Forward - Backward or Baum-Welch are commonly used to locally optimize HMM parameters (i.e., observation and transition probabilities). In this paper we propose a general approach based on Genetic Algorithms (GAs) to evolve HMM with Gaussian mixture densities. The problem appears when experts assign probability values for HMM, they use only some limited inputs. The assigned probability values might not be accurate to serve in other cases related to the same domain. We introduce an approach based on GAs to find out the suitable probability values for the HMM to be mostly correct in more cases than what have been used to assign the probability values. For this purpose, a sample database containing speech files of Algerian speakers is used.
机译:现在,语音识别应用程序变得越来越有用。 在自动语音识别(ASR)系统中,隐藏的马尔可夫模型(HMMS)已被广泛用于建模时间语音信号。 诸如前后或BAUM-WELCH的迭代算法通常用于局部优化HMM参数(即观察和过渡概率)。 在本文中,我们提出了一种基于遗传算法(气体)的一般方法,以通过高斯混合密度进化HMM。 当专家分配HMM的概率值时,出现问题,它们只使用一些有限的输入。 在与同一域相关的其他情况下,分配的概率值可能不准确。 我们介绍了一种基于气体的方法,以便在更多的情况下发现HMM的合适概率值大多是正确的,而不是用于分配概率值的情况。 为此目的,使用包含阿尔及利亚扬声器的语音文件的示例数据库。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号