基于书写摩擦力的签名识别方法

由于动态测量坐标系的存在, 使得数字笔在不同的握笔姿态下, 引起书写平面内摩擦力的分解不同. 由此本文提出采用书写摩擦力进行在线签名识别. 该方法将摩擦力的方向在直角坐标系内进行特征量表示, 并与三维书写力的大小相结合, 实现概率值和距离测度信度级的融合. 在进行真假签名的判别中, 加入快速剔除的方法, 进行两级判别, 以提高识别效率. 实验结果表明, 基于签名书写力大小和方向特征的融合识别方法互相补充, 相对于单个方法具有更高的识别率, 系统性能得到显著改善.

Signature Verification Using Handwriting Friction Force

Dynamic coordinates exist in handwriting input equipment of the digital pen. Consequently, various holding gestures of the digital pen lead to different decomposition ways of friction force. An approach using handwriting friction force for online signature verification is presented. The feature representation of friction direction is described in Cartesian coordinates. A novel system combines the extracted signature features with the characteristic of force magnitude to realize fusion of distance measures and probability. A quickly removed method is applied to the proposed system to improve the recognition efficiency. Two-stage discrimination is executed in distinguishing true and false signatures. Experiment result explores that fusion of handwriting force magnitude and direction for signature verification is shown to give complementary recognition information. As opposed to a single method, the fusion approach has a much higher recognition rate.