Soft-output Bayesian decoding and computation-MSE-oriented modulation design

Develop soft-output Bayesian decoding for HiCoMAC and modulation-design methods that directly target computation mean squared error (MSE), rather than relying on sequence-oriented Viterbi decoding and the computational free-distance criterion.

Background

The paper notes that the histogram-state trellis can support soft-output forward–backward inference, such as the BCJR algorithm, to obtain posterior marginals of the arithmetic-sum symbols. Because the quantized computation value is linear in these symbols, posterior means could be used to form a Bayesian MMSE estimate of the quantized sum.

The proposed work instead focuses on sequence-oriented Viterbi decoding and uses computational free distance as the modulation-design objective. The unresolved extension is therefore twofold: develop Bayesian soft-output decoding for the arithmetic-sum computation and design modulation directly for computation MSE rather than for sequence distinguishability.