A new method for blind identification of multipath channel by exploiting signal cyclostationarity

This paper proposes a new blind identification algorithm for identifying an unknown multipath channel driven by cyclostationary inputs. Like many other blind identification algorithms, this method also exploits only the second-order cyclic statistics of digital modulated signals, but it further makes use of other useful a priori known information such as the baseband pulse shape and the modulation type. In light of the fundamental relationship between the spectral correlation densities (SCD) of channel input and output, we apply some simple techniques of harmonic retrieval, outlier censoring, and combinatorial analysis to determine the number of paths and estimate the parameters, i.e., the path delays and attenuation For different baseband pulse shapes such as rectangular and raise-cosine pulses, some practical discussions on the channel identifiability and resolution performance of the algorithm are included. Simulation results are also provided to demonstrate its performance in the presence of stationary noise and interfering signal.