A Novel Linear Mixing Information Confusing Framework Based on Minimum Nongaussianity Estimation

摘要：

Maximum nongaussianity estimation is now a popular method in independent component analysis (ICA), in which the nongaussianity of outputs is maximized and then independent components are recovered. In this paper, from the inverse viewpoint of ICA, a novel linear mixing information confusing framework is presented based on the theory of minimum nongaussianity estimation (MNE). For given independent nongaussian signals to be transmitted, a linear mixing system is introduced in information transmitting system before information encoding, while the linear vector or matrix is decided by minimizing the nongaussianity of mixtures, e.g. the kurtosis of mixtures, thus nearly Gaussian distributed signal can be obtained and meanwhile the information can are mutually confused and hidden. In the receiver terminal, independent sources can be recovered via ICA algorithms. Simulation results based on speech signals confusing show that the presented method is simple and effective.

关键词： Independent Component Analysis Maximum Nongaussianity Estimation MinimumNongaussianity Estimation Kurtosis

作者: Gang Wang Hongwei Li Fangyu Wang

作者单位: Telecommunication Engineering Institute, Air Force Engineering University, Shanxi, 710077, China;Uni Telecommunication Engineering Institute, Air Force Engineering University, Shanxi, 710077, China

会议类型: 国际会议

会议名称: 第七届国际测试技术研讨会

会议地点: 北京

会议语种:英文

在线出版日期: 2007-08-05（万方平台首次上网日期，不代表论文的发表时间）