Dual mode decision feedback equalization scheme under parallel adaptation

In this paper, we extend the fast convergence blind decision feedback equalization (DFE) scheme developed by Labat, Macchi, and Laot in 1998 so that fast convergence is also possible for larger constellations and noisier conditions. The original scheme starts up with a linear equalizer (LE) and switches to a DFE after the initial convergence has been achieved. The switching criterion between modes is based on the decision directed mean-squared-error. Our proposed scheme simultaneously adapts the LE and the DFE in a parallel fashion such that the switch-over between modes is no longer necessary. Consequently, smoother and possibly faster convergence is achieved as switching disruptions are reduced. The combination techniques are adopted from dual-mode algorithms found in the literature.