Reduced Complexity Affine Projection Algorithm Based on Variable Projection Order and Multiple Sub Filter Approach

Abstract

The main aim of the paper is to reduce the complexity of affine projection algorithm (APA) by using multiple sub filter together with variable step size and projection order so that there is no compromise in the performance measure. Here the entire filter length is segmented into smaller segments (sub filter) in order to reduce the projection order which is the main cause for the complexity. However the steady state error increases with the use of the multiple sub filter approach. To maintain an overall lower steady state error, a variable step size together with a variable projection order is used. A simple rule for the automatic change of projection order based on the instantaneous value of output error is also proposed. Therefore faster convergence, lesser steady state error criteria is met with the reduction in computational complexity and computational time. Simulation is performed in the context of echo cancellation application to prove the validity of the algorithm.