Blind identification of sparse multipath channels using cyclostationary statistics

Blind identification of a wireless communication channel is an important issue in communication system design. Most existing blind system identification techniques process the unknown information of the system from its output only. However, in many practical situation partial knowledge of the system transfer function is available. By relying on this known information, the performance of channel identification and equalization can be significantly enhanced. In this paper, we introduce a new system identification technique that exploits both the a priori knowledge of the pulse shape filter and the multi-path channel propagation model. The approach consists first in processing the cyclo-spectrum of the system output that is shown to be superimposed exponential function of the channel propagation delays and attenuations. Then, the frequency parameters, i.e., channel propagation parameters, are later estimated using the Matrix Pencils (MP) frequency estimation method [7].