Abstract
A novel equalization algorithm utilizing improper nature of the intersymbol interference (ISI) is introduced in this paper. We show that full exploitation of the available information on the second-order statistics of the observed signal entails widely linear processing and that previously known linear minimum mean square error (MMSE) equalizers represent sub-optimum solutions. The proposed scheme is generally applicable for both real and complex signal constellations. The results show that accounting for the improper nature of the ISI leads to significant performance gain compared to conventional equalization schemes.