A way to do off-line handwriting recognition is to generate an equivalent on-line signal of a letter image and to use an on-line recognition system to recognize this letter. To do this, the drawing order has to be recovered using handwriting knowledge. Our approach to recover the drawing order consists in proposing several starting and ending points. Several paths are generated and the best one is chosen. This paper presents the handwriting knowledge introduced at several steps of our method and its contribution for the recognition process. Experimentations have been carried out on isolated lower case multistroke letters from a database including both on-line and off-line signals. An on-line system has been used to recognize these letters.