一种信号数目估计的小波包分解优化算法  

Optimizing Algorithm in Determining the Number of Signals Using Wavelet Packet Decomposition

作　　者：汪晋宽[1] 薛延波[1] 刘志刚[1] 刘福来[1] 

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110004

出　　处：《东北大学学报（自然科学版）》2003年第8期743-746,共4页Journal of Northeastern University(Natural Science)

基　　金：教育部科学技术研究重点项目(02085)

摘　　要：分析了入射信号数目估计的信息理论准则·针对在低信噪比以及数据快照数目较小的情况下,Akaike信息准则(AIC)和最小描述长度MDL都不能正确估计入射信号数目的问题,提出了采用小波包分解方法对MDL准则进行优化的算法,称之为wpMDL算法·考虑子带分解具有预测偏差小和模式隔离等特点,采用基于一维小波包分解的'伪二维'子带分解算法,不仅有效地改善了MDL准则的估计性能,而且计算量比二维分解降低了·仿真结果验证了提案算法的有效性,无论是对信噪比的变化还是数据快照数目的变化,wpMDL算法都能正确地估计入射信号的数目·Detection of number of signals by information theoretic criteria is given. In the occasions of low signaltonoise ratio (SNR) and large snapshots,incorrect estimate of signal number is often yielded with both Akaike Information Criteria (AIC) and Minimum Description Length (MDL). A wavelet packet decomposition (WPD) based algorithm is proposed to optimize MDL, which is addressed wpMDL. With the features of minimum prediction error and mode isoloation, pseudo 2D subband decomposition algorithm based on onedimension (1D) wavelet packet decomposition is introduced to not only improve the estimate performance but also save the computation by O(N4) compared with 2D WPD.

关 键 词：信号数目 Akaike信息准则 最小描述长度 小波包 子带分解 

分 类 号：TN911.7[电子电信—通信与信息系统]