In this study, we present a novel approach to measuring the similarities for an on-line Chinese signature verification system. The proposed system comprises two stages, the learning stage and the verification stage. The former generates all the reference signatures for the users, and the latter calculates the similarities between the test signature and the corresponding reference signature. A two-step matching is proposed that includes coarse matching and fine matching. The method developed in [11] is adopted in this study to accomplish the coarse matching. However, the similarity between the test and reference signatures obtained at the first step is not accurate enough. To achieve a better result, the method developed in [12] is modified in this study to measure the similarity of the two signatures by transforming the corresponding signature strokes into the frequency domain for further minor adjustment. The experimental results reveal the practicability of the proposed methods.
展开▼