Chain Code Extraction of Handwritten Recognition using Particle Swarm Optimization

As one of soft computing optimization tools, Particle Swarm Optimization (PSO) has been applied in many fields such as in handwritten recognition and classification. Associated with the development of PSO algorithm is the data representation to be used as input to the algorithm. One of data representation scheme is Freeman chain code (FCC). As one of traditional scheme in representing data, FCC is still relevant as a scheme in data compression and representation. The scheme is proposed in this paper to represent Latin handwritten as isolated character and as input to the PSO algorithm. The main problem related to FCC and handwritten recognition to be solved in this paper is in representing and recognizing character because the length of the FCC depends on the starting points. Therefore, every pixel (or node) must be coded with a specific number depending on its direction. In addition to the isolated characters especially the uppercase characters, the traversing process of each pixel (or node) of this type is more difficult because of the problem in finding several branches and revisiting the same nodes. As a solution, one continuous route is proposed to solve the problems which cover all of the nodes in the character image. The proposed PSO algorithms extract the FCC from the thinned binary image of the character and recognize difficult characters based on the series of FCC. Two performance parameters used to measure the performance of the PSO algorithm are the computational time and route length. Result shows that the PSO algorithms successfully extract FCC and recognize character at a relatively shorter computational time and route length.