Two variable step-size adaptive algorithms for non-Gaussian interference environment using fractionally lower-order moment minimization

• Published in: Digital signal processing Vol. 23; no. 3; pp. 831 - 844
• Main Authors: Zheng, Yahong Rosa, Nascimento, Vítor H.
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.05.2013
• Subjects: Adaptive algorithms; Algorithms; Bernoulli–Gaussian distribution; Compound K distribution; Errors; Fractionally lower-order moment (FLOM) algorithm; Gaussian; Interference; Least mean p-moment (LMP) algorithm; Minimization; Non-Gaussian; Non-Gaussian interference suppression; Optimization; Robust adaptive filter; Variable step size; Variable step size; Compound K distribution; Non-Gaussian interference suppression; Least mean p-moment (LMP) algorithm; Robust adaptive filter; Fractionally lower-order moment (FLOM) algorithm; Bernoulli–Gaussian distribution
• ISSN: 1051-2004; 1095-4333
• DOI: 10.1016/j.dsp.2012.12.019

Two variable step-size adaptive algorithms using fractionally lower-order moment minimization are proposed for system identification in non-Gaussian interference environment. The two algorithms automatically adjust their step sizes and adapt the weight vector by minimizing the p-th moment of the a posteriori error, where p is the order with 1⩽p⩽2, thus they are named as variable step-size normalized least mean p-th norm (VSS-NLMP) algorithms. The proposed adaptive VSS-NLMP algorithms are applied to both real- and complex-valued systems using low-complexity time-averaging estimation of the lower-order moments. Simulation results show that the misalignment of the proposed VSS-NLMP algorithms with a smaller p converges faster and achieves lower steady-state error in impulsive interference and/or colored input environment. The adaptive VSS-NLMP algorithms also perform better than the adaptive fixed step-size (FSS) NLMP in both Gaussian and finite-variance impulsive interference environments. A theoretical model for the steady-state excess mean-square error is also provided for both Gaussian and Bernoulli–Gaussian interference.