Abstarct: In this study, a method for online Persian handwritten recognition, baxsed on the recognition of sub-words’ composing letters is proposed. The input of systems for online recognition is a set of points related to the handwritten which are generally gathered by handheld digital systems like tablets and smart phone at the time user writes on the screen by touch or electronic pen. So this is called online. In proposed method in a simplification step, the input data are represented as a set of horizontal and vertical lines with two direction and magnitude vector that these vectors are actually the extracted features of the data. Then reference patterns related to alphabets on various parts of subwords are extracted from feature vectors to create the knowledge baxse. In the recognition step, a list of subwords are suggested containing the composing letters recognized by template matching technique with a proposed dynamic programming algorithm. And finally in post-processing step, we filter the suggestions that are not available in dictionary or are not in agreement with the input data in terms of number and or position of sub-strokes. by the simplification and the proposed algorithm to calculate the edit distance, The success percent of recognition of Persian isolate letters is 95.2% and the percent for the sub words is 63.29% in the first three suggested subwords. Although the success percent for this method is lower than the holistic approaches but the proposed method has a main advantage over the holistic methods in recognition of sub-words composing letters. To recognize the new sub-words by the proposed method, we just need to add them to the system’s dictionary, but for hoslistic methods we require new samples of handwritten.