Constraint Natural Gradient Algorithm with non adaptive metric

This paper presents a critical augmentation to the previously developed Constraint Adaptive Natural Gradient Algorithm (CANA), by incorporating a non adaptive metric in place of adaptive Riemannian metric. Inclusion of this fixed metric significantly reduces the complexity of previously developed technique. Complemented with Constrained optimization techniques the algorithm is capable of rapidly adjusting the response of an array of sensors to a signal coming from Direction of Interest (DoI) and suppressing noises coming from other directions. Mathematical analysis and realistic MATLAB simulations are developed to underpin the fast convergence of the adaptive weights, and the suitability of algorithm for operations in adverse environments. The technique has been further extended to the variable step size Adaptive Array Processing..