Work

RR-DCD: A Recursive-Residual DCD-Based Low-Complexity Variable Forgetting Factor Scheme for Sparse PRLS-DCD Algorithms

Year: 2026
Type: article
Abstract: Sparse and long-delay time-varying channel impulse responses are challenging for underwater acoustic (UWA) transmission and make low-complexity adaptive channel estimation and equalization essential for computationally constrained receivers. Proportionate recursive least-squares (PRLS) and dichotomous coordinate descent (DCD) iterations improve sparse-system adaptation while reducing the computational cost of coefficient updates. However, in current PRLS-DCD algorithms, the existing covariance-aware approaches of variable forgetting factors (VFFs) still suffer from considerably high complexity, reaching O(N2), which can dominate the overall implementation. To further reduce the computational complexity of VFF implementation, we propose a recursive-residual DCD (RR-DCD) VFF scheme, which replaces the exact quadratic-form evaluation (Exact-q) with a warm-started auxiliary-solver strategy. The previous solution is propagated, the initial residual is computed using vector operations, and a fixed number of DCD corrections is performed to refine the solution prior to VFF mapping. We further derive a residual-dependent error bound and the resultant computational complexity. Specifically, the proposed RR-DCD module for computing the VFF statistic achieves O(N) complexity with a fixed number of auxiliary iterations. Paired Monte-Carlo experiments on abruptly varying sparse channels and the UWA channel generated using Bellhop and field-collected sound speed profile are conducted to compare the proposed scheme against the Exact-q VFF and baselines. The results validate a favorable accuracy–complexity trade-off and deliver an effective low-complexity adaptive strategy for long-delay sparse-channel scenarios.
Language: English
Open Access status: gold
Estimated APC paid: $3,137
OpenAlex ID: W7235435784
FWCI:
Cites: 34
Cited by:
Related to: 10