Writer Identification Method using Inter and Intra-Stroke Information

摘要

Most research in the field of writer identification currently uses positional data to identify the writer. But there is no writer identification method that is suitable for practical use so far. So, we newly devise two algorithms to raise accuracy rate and find some parameters that have beneficial effects on writer identification. One algorithm is "Block Type Model" which is a method to analyze positional relation and shapes of Chinese characters. The other algorithm is "Hidden feature analysis" which is an algorithm that identifies the writer using multi-parameters with all the strokes. As a result, any eight Chinese characters were enough to achieve over 99.9% accuracy rate. Additionally, we discovered that there are personal attributes in the speed of each stroke. With the new algorithm, we raised accuracy rate and found out that parameters have beneficial effects on writer identification.