|
|
A Numerically Robust and Low-complexity Method of Signal Subspace Estimation |
Zhuang Xue-bin Lu Ming-quan Feng Zhen-ming |
Department of Electronic Engineering, Tsinghua University, Beijing 100084, China |
|
|
Abstract A numerically robust and low-complexity method of signal subspace estimation is proposed in the paper. The transform matrix to tridiagonalize the covariance matrix of observation data is constructed in the forward recursion of multistage Wiener filter (MSWF), and its columns span the signal subspace. Compared with the traditional method of correlation subtractive structure, the forward recursion in the method is implemented with the Householder unitary transform. Therefore, it strengthens significantly the orthogonality of basis vectors in the signal subspace and improves the numerical robustness, especially in the finite-precision implementation. Besides, a method of calculating the transform matrix is proposed to reduce the computational complexity based on the unitary property of the Householder matrix and backward accumulation of matrices. Finally, simulation results demonstrate the numerical robustness and computational efficiency of the proposed method.
|
Received: 29 October 2009
|
|
Corresponding Authors:
Zhuang Xue-bin
E-mail: zhuangxb06@mails.tsinghua.edu.cn
|
|
|
|
|
|
|