A STATISTICAL SUBSPACE METHOD FOR BLIND CHANNEL IDENTIFICATION IN OFDM COMMUNICATIONS

This paper presents a subspace method for bland channel es-t imation based on a special output correlation matrix. T h e approach relies on the i.i.d. assumption of the data sequence and uses the cyclic prefix redundancy present in OFDM sys-tems or single-carrier systems with frequency domain equal-ization. This method has an important feature that allows channels to be longer than the cyclic prefix. I n addition, un-known frame synchronization can be accommodated. There is no constraint on the zero locations of the channel and the performance is asymptotacally independent of white noise. The method is compared with a simple correlation approach and a deterministic subspace method. 1