Dynamic programming matching applied to off-line handwritten character recognition

Through adding a smoothness term to the Euclidean distance, and removing the restriction for the conventional DP matching method that the matching function must be a monotonously increasing, we extend the DP matching method, which is often used to calculate the distance between two on-line handwritten characters, to the one which can be used to calculate the distance between an on-line handwritten character and an off-line handwritten character. When given a set of on-line handwritten characters as prototypes, we can recognize an unknown off-line handwritten character by use of the proposed extended DP matching method, together with finding one of the possible Eulerian paths for the off-line handwritten character image. The experimental studies have revealed the effectiveness of our method