Combination of affine projection sign algorithms for robust adaptive filtering in non-gaussian impulsive interference

摘要

A method for minimising the L₁ norm relying on the convex combination of two affine projection sign algorithms is proposed. The major drawbacks of the affine projection sign type algorithms are a poor convergence rate and instability in time-variant scenarios, especially in the presence of abrupt changes. A convex combination method is used to obtain a superior performance. Instead of minimising the squared error in the conventional convex combination theory, the minimisation of the L₁ norm is introduced to offer a more robust solution in the presence of a non-Gaussian impulsive interference. The good performances in terms of the convergence rate and the steady-state error are demonstrated in a plant identification experiment that includes the impulsive noise and abrupt changes.