Online writeprint identification via Multi-PRM

To deal with the high dimensionality and redundancy of the online writeprint, this paper proposed an ensemble learning approach based on Multiple Probabilistic Reasoning Model. An inverse method of pseudo-random number generator is employed to construct multiple random subspaces, and then the base classifier is trained in each subspace. Finally, each classifier is aggregated to construct a strong ensemble through a combination strategy. The experiment is conducted on a real dataset, focusing on the approach´s parameters, sampling rate and granularity of space dividing. The results show that the proposed method is effective and appropriate values of parameters can effectively improve the identification performance of online writeprint.