Low-Complexity Double EP Approximation for Parallel Detection-Decoding in MIMO Systems

In this letter, we propose new expectation propagation (EP)-based iterative detection and decoding (IDD) algorithms. The proposed algorithms can replace all exponential operations and eliminate the matrix inversions in both inner and outer loops, reducing significantly the computational complexity. In addition to complexity reduction, the proposed algorithm decouples the data dependency between inner loops and outer loops. Consequently, we propose the parallel EP-IDD framework to enable simultaneous iterations of detection and decoding. Simulation results demonstrate that the proposed algorithms can outperform the minimum mean-squared error (MMSE)-based IDD algorithm by 0.55 dB at the bit error rate of 10-3 with less computational complexity consumption. Compared with the state-of-the-art EP-based IDD algorithms, the proposed algorithm achieves a 69.55% to 95.27% reduction in computational complexity without noticeable performance loss.