Low-Complexity Normalized Subband Adaptive Filter for the Censored Regression Model

The performance of the normalized subband adaptive filtering (NSAF) algorithm will be degraded when processing signals with missing data. In this brief, a normalized subband adaptive filtering algorithm based on censored regression (CR-NSAF) is proposed, which is derived by using the conditional expectation of the desired signal. However, as a traditional subband algorithm, the CR-NSAF algorithm still suffers from the problem that the computational complexity increases as the number of subbands increases. Thus, a low-complexity CR-NSAF algorithm (SCR-NSAF) is proposed, which reduces the computational complexity by using the subband selection method. The simulation results and analysis show that the proposed algorithms achieve superior convergence performance in comparison to the traditional algorithm under the censored regression (CR) environment.