Novel robust filtering algorithms employing generalized correntropy within p-norm loss

Abstract

The least mean $p$-power algorithm (LMPA) has been widely applied in signal processing. However, with large impulsive noise, LMPA exhibits low convergence speed. In this paper, to enhance the filtering performance, we develop new adaptive filtering algorithms, which consist of three parts. Firstly, a novel cost function is presented by incorporating a generalized correntropy within $p$-norm loss, mitigating the impact of impulsive noise. Subsequently, to improve the convergence rate, revised least mean $p$-norm algorithms (RLMPAs) based on the novel cost function are proposed. Then, the stability performance is analyzed. Simulations on channel estimation and echo cancellation show that the proposed algorithm performs much better than generalized maximum correntropy criterion algorithm in terms of the convergence rate.