Squared Sine Adaptive Algorithm and Its Performance Analysis.

The squared sine adaptive (SSA) algorithm is presented for identification scenarios, such as acoustic-echo cancellation (AEC) applications, in non-Gaussian environments. To devise the SSA algorithm, a novel cost function is constructed by exerting a sliding window-type squared sine function on the estimation error vector, which provides robustness in impulsive-noise environments and speeds up convergence when the input is colored. Theoretical results are presented for predicting the mean-weight, convergence, transient excess-mean-square-error (EMSE), and tracking behaviour. Moreover, the minimum EMSE and the optimum step size for tracking are presented. The computational complexity of the SSA algorithm has also been investigated. Numerical experiments demonstrate that results of the theoretical analysis match the simulated results very well and the proposed SSA algorithm outperforms known algorithms in AEC applications.