• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>International Journal of Innovative Computing Information and Control >SVM-EMBEDDED FLCMAP SPEAKER ADAPTATION USING A SUPPORT VECTOR MACHINE TO IMPROVE FUZZY CONTROLLERS OF FCMAP
【24h】

SVM-EMBEDDED FLCMAP SPEAKER ADAPTATION USING A SUPPORT VECTOR MACHINE TO IMPROVE FUZZY CONTROLLERS OF FCMAP

机译：使用支持向量机改进FCMAP的模糊控制器的SVM嵌入式FLCMAP扬声器自适应

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Maximum a posteriori (MAP) adaptation is currently one of the most widely used speaker adaptation techniques. However, the effectiveness of MAP relies significantly on the quality and quantity of adaptation data. Although an FCMAP approach that incorporates a fuzzy logic controller (FLC) into MAP could avoid performance degradation that is caused by a poor MAP estimate resulting from relatively small amounts of adaptation data, the performance of FCMAP cannot be maintained in a situation where the quality of the acquired adaptation data is low. To address the issue of FCMAP, this research proposes a support vector machine (SVM)-embedded FLCMAP method that improves the fuzzy controller in the conventional FCMAP. The developed SVM-embedded FLCMAP determines the quality of adaptation data by using an SVM. The SVM-estimated data quality information, together with the data quantity, can then be used to drive the fuzzy inference operations of the FLC. The proposed SVM-embedded FLCMAP provides a method to ensure the robustness of FCMAP against data inferiority. The experimental results show that an SVM-embedded FLCMAP outperforms FCMAP, especially when encountering substandard data. The proposed SVM-embedded FLCMAP performs more efficiently than hybrid SVM-FLC MAP and is more flexible for adaptive data utilization.
机译：当前，最大后验（MAP）适应是最广泛使用的说话者适应技术之一。但是，MAP的有效性在很大程度上取决于适应数据的质量和数量。尽管将MAP包含模糊逻辑控制器（FLC）的FCMAP方法可以避免由于适应数据量相对较小而导致的MAP估计不佳而导致性能下降，但是在以下情况下无法保持FCMAP的性能：所获取的适配数据低。为了解决FCMAP的问题，本研究提出了一种嵌入支持向量机（SVM）的FLCMAP方法，该方法改进了传统FCMAP中的模糊控制器。所开发的嵌入了SVM的FLCMAP通过使用SVM来确定自适应数据的质量。然后，可以使用SVM估计的数据质量信息以及数据量来驱动FLC的模糊推理操作。所提出的嵌入SVM的FLCMAP提供了一种确保FCMAP抵抗数据劣势的鲁棒性的方法。实验结果表明，嵌入SVM的FLCMAP优于FCMAP，尤其是遇到不合格数据时。所提出的嵌入SVM的FLCMAP比混合SVM-FLC MAP更有效，并且对于自适应数据利用更灵活。

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