Blind identification using channel length estimation: subspace approach based on CGM

摘要

Subspace methods (SSM) are an effective approach for blind channel identification. However, these methods have two major disadvantages: i) They require a large amount of computation for the eigen-value decomposition (EVD) and the singular-value decomposition (SVD), and ii) they require the prior knowledge of accurate channel length. In this paper, we discuss new algorithm for blind channel identification using the conception of principal component analysis (PCA), which is based on the orthogonality between the subspaces spanned by the column vectors of the impulse response matrix (the impulse response subspace) and the noise-subspace, and using the property of conjugate gradient method (CGM). The new algorithms do not need to calculate both EVD and SVD, and to get the prior knowledge of accurate channel length. Furthermore, the new blind algorithm has computations O(m{sup}2) where m is the data vector length.