A Design of Information Theoretic Criteria for Detecting the Number of Incoherent Signal

摘要：

This paper proposes a novel design method of criteria for detecting the number of signals superimposed in multichannel time-series. Based on probabilistic properties of difference in maximum log likelihood at infinite SNR, penalty functions in information theoretic criteria are designed by giving specific upper bounds of error probabilities. The proposed design method uses an approximation of probability distribution functions of the difference in maximum log likelihood are approximated. Finally, simulation results are shown to demonstrate flexible criteria for detecting the number of signals can be designed in the case that the number of available samples of observation vectors is small and also large.

关键词： detection MDL ML information theoretic criteria small samples

作者: MingZhang Haihua Chen Tingting Teng Masakiyo Suzuki

作者单位: Kitami Institute of Technology, Japan

会议类型: 国际会议

会议名称: 2006现代科技国际研讨会(The International Workshop on Modern Science and Technology in 2006)

会议地点: 北京

会议语种:英文

页码: 358-365

在线出版日期: 2006-04-01（万方平台首次上网日期，不代表论文的发表时间）