On channel identifiability based on second order cyclic spectrum

The author presents results that prove the identifiability of linear finite-dimensional channels with higher than minimum bandwidth based on the use of second-order cyclic statistics only. Since the sampled channel output is also cyclostationary, second-order cyclic statistics of data samples obtained at higher than the Nyquist sampling rate can be used to identify the unknown linear channel. Without the need for higher-order statistics, fractionally spaced equalizers exploiting the cyclostationary property of the channel output may become effective alternatives to conventional schemes based on higher-order statistics