Blind channel estimation via subspace approximation

Subspace channel estimation technique has been well studied. It relies on subspace decomposition on the data covariance matrix to identify either signal subspace or noise subspace. However, the subspace decomposition process is computationally expensive. This paper applies a novel subspace approximation (SA) idea to bypass this process for channel estimation. It is based on a recently proposed "power of R" (POR) multiuser detection technique that raises power of estimated data covariance matrix to a positive integer to approximate rather than estimate the noise subspace component. Thus complexity is significantly reduced while satisfactory performance is still maintained. Channel estimation performance is studied in detail and compared with that of the well-known subspace method in the literature.