Decision feedback blind symbol estimation by adaptive least squares smoothing

A decision feedback blind symbol estimation algorithm based on the least squares smoothing approach is proposed for single-input multiple-output finite impulse response systems. With the finite alphabet property, the input signal is estimated based on the past detected symbols and the least squares smoothing error of the observation. Implemented both time and order recursively, the proposed algorithm is adaptive to channel variation and has low complexity both in computation and in VLSI implementation. Based on a deterministic model, this algorithm has the finite sample convergence property, i.e., the input signal can be perfectly detected with a small set of data samples in the absence of noise.