...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Journal of Computers >Extraction of Unique Independent Components for Nonlinear Mixture of Sources
【24h】

Extraction of Unique Independent Components for Nonlinear Mixture of Sources

机译：用于源非线性混合的独特独立组分的提取

获取原文
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

—In this paper, a neural network solution toextract independent components from nonlinearly mixedsignals is proposed. Firstly, a structurally constrainedmixing model is introduced to extend the recently proposedmono-nonlinearity mixing model, allowing that differentnonlinear distortion are applied to each source signal. Basedon this nonlinear mixing model, a novel demixing systemcharacterized by polynomial neural network is thenproposed for recovering the original sources. The parameterlearning algorithm is derived mathematically based on theminimum mutual information principle. It is shown thatunique extraction of independent components can beachieved by optimizing the mutual information cost functionunder both model structure and signal constraints. In thisframework, the theory of series reversion is developed withthe aim to perform dual optimization on the polynomials ofthe proposed demixing system. Finally, simulation resultsare presented to verify the efficacy of the proposedapproach.
机译：- 本文提出了一种从非线性混合标记的神经网络解决方案倾斜独立组分。首先，引入了结构上约束的混合模型以扩展最近的预设常规非线性混合模型，允许将不同的环路线圈施加到每个源信号。基于该非线性混合模型，通过多项式神经网络进行了一种新的解映射系统，用于恢复原始来源。参数图形算法基于最大的相互信息原理在数学上衍生。显示通过优化模型结构和信号约束中的相互信息成本功能，可以靠近独立组件的单一提取。在此，开发了串联逆变理论，目的是对所提出的解混系统的多项式进行双重优化。最后，仿真结果显示验证了拟议的效果。

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