Prediction based direct blind decision feedback equalization

We propose a blind decision feedback equalizer (DFE) using second-order statistics (SOS). The proposed blind decision feedback equalization technique does not require channel estimation and obtains the equalizer coefficients directly from SOS of the received signals. In the case of MIMO channels, autoregressive (AR) process is also moving average (MA) process. Therefore, the output of the optimized multistep prediction error filter (PEF) can be represented as a product of channel partial impulse response and transmitted sequence. It is shown that a backward multistep FEF can be used as the feedforward filter (FFF) of blind DFE. The corresponding feedback filter is obtained from the symbol-rate partial channel impulse response. The proposed algorithm has several advantages which is not shared with existing blind channel estimation techniques. The proposed DFE achieves the stable performance without the exact channel order estimation. The performance of the equalizer can be improved by the nonlinearity of the decision device which alleviates the inaccuracy of the multistep PEF.