Complexity-Reduced NOMP for OFDM: Heuristic Stopping and Approximate Updates

Channel estimation is important for Orthogonal Frequency-Division Multiplexing (OFDM) in wireless channel communication and requires algorithms that offer the best accuracy while at the same time have very low computational and runtime complexities. Newtonised Orthogonal Matching Pursuit (NOMP) is a promising algorithm for channel estimation; however, it suffers from high computational complexity due to repeated refinement and least-squares updates. In this paper, we propose a low complexity NOMP variant that reduces the dominant computational cost through three modifications: (i) a residual energy-based stopping criterion for NOMP to avoid expensive CFAR evaluation, (ii) a partial cyclic refinement with frozen atoms, and (iii) approximate one-sweep per atom least-squares updates. Complexity analysis shows a reduction from O(K³) to O(KN) in the gain update and from O(K²N) to O(KN) in refinement. Simulation results show that the proposed method achieves ∼87% reduction in runtime, while the symbol error rate (SER) performance is comparable to classical NOMP and outperforms Oversampled OMP at high signal-to-noise ratio (SNR). These results show that NOMP can be computationally efficient for OFDM systems without sacrificing estimation accuracy.