A novel quasi accelerated over-relaxation method for efficient signal detection

This letter proposes a novel and efficient quasi accelerated over-relaxation (QAOR) based signal detector that significantly reduces complexity caused by the matrix inversion of minimum mean square error (MMSE). QAOR introduces a relaxation parameter to the diagonal elements of the matrix to increase its dominance and boost numerical stability for faster convergence in the large systems compared to the standard accelerated over-relaxation (AOR) method. To further improve performance, an adaptive method is proposed for selecting acceleration and relaxation parameters. In addition, a novel initialization method is introduced to improve overall detection accuracy and performance. Experimental results and complexity analysis demonstrate that the proposed method outperforms existing AOR and recently developed successive over-relaxation (SOR) methods while achieving near MMSE performance. These improvements highlight the proposed algorithm as an efficient and reliable solution for signal detection in massive MIMO systems.