Annealed SOR-Based Gibbs Sampler for MIMO Detection in Frequency-Selective Channels
Zhiheng Zhang, Le Yang, Jun Tao, Yan Huang · IEEE Signal Processing Letters · 2024
Asuccessive over-relaxation (SOR) based Gibbs sampler has been proposed for Markov chain Monte Carlo (MCMC) symbol detection in multiple-input multiple-output (MIMO) communication systems. It outperforms the conventional standard Gibbs sampler with even faster convergence. Such an advantage, however, is achieved only in the high signal-to-noise ratio (SNR) region. This paper proposes an annealed version of the SOR-based Gibbs sampler, not only expanding the working SNR range significantly but also improving the detection performance. Besides, we extend the MCMC symbol detection originally developed under frequency-flat channels to the more general case with frequency-selective channels. Numerical results corroborate the superiority of the proposed detection scheme.