Non-Parametric VSS-NLMS Algorithm With Control Parameter Based on the Error Correlation

The behavior of variable step-size least-mean-square (VSSLMS) algorithms is strongly affected by measurement noise. Thereby, aiming to maintain an adequate performance of these algorithms, their parameters must be adjusted whenever changes occur in the signal-to-noise ratio (SNR) of the adaptive system. A well-known VSSLMS algorithm based on error correlation provides a performance enhancement for low SNR environment; however, its immunity to measurement noise changes is still poor. This paper presents a new nonparametric variable step-size normalized LMS (VSS-NLMS) algorithm with control parameter based on error correlation. This approach is very robust to measurement noise changes, not requiring any manual adjustment of algorithm parameters. Numerical simulation results confirm the effectiveness of the proposed algorithm, considering a system identification problem.