New autocorrelation decomposition method for robust channel estimation with unknown channel order

A new subspace method based on autocorrelation decomposition is developed for channel estimation. It is shown that under suitable conditions, identifiability can be guaranteed in the presence of channel order overestimates, in a clear contrast to most subspace methods such as [E. Moulines et al., Feb. 1995]. Moreover, simulations show that the proposed scheme is very robust against channel order overestimates, thereby de-emphasizing the need for complex order estimation procedures, i.e., a conservative order (over)estimate based on empirical knowledge such as maximum delay spread can be used in practice.