Probabilistic Equalization With a Smoothing Expectation Propagation Approach

In this paper, we face the soft equalization of channels with inter-symbol interference for large constellation sizes, M. In this scenario, the optimal BCJR solution and most of their approximations are intractable, as the number of states they track grows fast with M. We present a probabilistic equalizer to approximate the posterior distributions of the transmitted symbols using the expectation propagation (EP) algorithm. The solution is presented as a recursive sliding window approach to ensure that the computational complexity is linear with the length of the frame. The estimations can be further improved with a forward-backward approach. This novel soft equalizer, denoted as smoothing EP (SEP), is also tested as a turbo equalizer, with a low-density parity-check (LDPC) channel decoder. The extensive results reported reveal remarkably good behavior of the SEP. In low dimensional cases, the bit error rate (BER) curves after decoding are closer than 1 dB from those of the BJCR, robust to the channel response. For large M, the SEP exhibits gains in the range of 3-5 dB compared to the linear minimum mean square error algorithm.

expectation propagation (ep); bcjr; low-complexity; turbo channel equalization; isi; turbo-equalization; intersymbol interference; sequence estimation; decision-feedback; linear codes; algorithm; channels; graphs

Probabilistic Equalization With a Smoothing Expectation Propagation Approach Articles