An improved global-best-driven flower pollination algorithm for optimal design of two-dimensional FIR filter

The design of two-dimensional (2D) digital filter is a higher-order, nonlinear, and multi-modal optimization problem. This paper presents an improved Global-best-driven Flower Pollination Algorithm, named as GFPA, for the design of 2D Finite Impulse Response (FIR) filters. Two methods have been proposed-the first method minimizes the weighted square error via GFPA and the second method finds the coefficients of one-dimensional FIR filter by GFPA before McClellan transformation. The performance of proposed algorithm has been compared with state-of-the-art algorithms and the simulation results show significant improvements. For the design of a 15x15 circular symmetric filter, an average reduction of 55% in fitness function evaluation and 72% in execution time is observed. Further, the experiment on CEC 2014 benchmark functions demonstrates better optimal solution than existing algorithms.