Blind equalization which only requires the output signal is of great interest in data communication systems when there is no available information data to train the equalizer. Nevertheless, complex sources are recovered with an arbitrary phase rotation in such applications. In this paper, we prove that blind equalization of complex source can be converted to blind separation of real 2×2 convolutive mixture where the source signals are the in-phase and quadrature components of complex source. As a result, multichannel blind deconvolution (MBD) algorithms can be used to address the problem. Simulations results show that the proposed scheme not only unravels the effects of convolution performed by a linear complex system operating on a complex input signals which have independent inphase and quadrature components, but also eliminate the phase indeterminacy simultaneously. What is more, the I/Q MBD scheme for blind equalization has more robust performance than the famous multimodulus algorithm (MMA).
Keywords-blind equalization; multichannel blind deconvolution (MBD); phase recovery.