Combined kurtosis driven variable step size adaptive line enhancer

Conventional LMS (least mean square) based adaptive line enhancer (ALE) has the disadvantages of slow convergence, low performance in suppressing non-Gaussian colored noise and in tracing time-varying signals. In order to overcome these defects greatly, a novel combined kurtosis driven variable step size LMS adaptive line enhancer (CKDALE) is suggested. In this algorithm, kurtosis definition is modified to handle non-Guassian noise to a great extent. The exponential type variable step size driven by both input signal kurtosis and error signal kurtosis is adopted and analyzed. It has demonstrated, by means of extensive simulations, that the proposed algorithm outperformed ALE in enhancing sinusoidal signals, suppressing non-Gaussian noise, and convergence rate, etc.