Proposed in this paper is a novel method of character segmentation for offline unconstrained handwritten Chinese text lines.In this method,first,a series of curved candidate segmentation paths are generated via a pre-segmentation algorithm.Then,the recognition information of isolated characters and the geometric information of text lines are integrated by using two modified quadratic discriminant functions,from which the confidence of the segmentation hypothesis is obtained.Moreover,the optimal segmentation hypothesis is searched by employing a dynamic programming algorithm.Finally,the proposed method is tested using 383 text lines in the HIT-MW database without the help of any language models.The segmentation accuracy reaches 89.70%,which means that the proposed method is effective.