User generic model for writer verification using multiband image scanner

Writer verification plays an important role in biometrics and forensics area. The familiar use of handwritten signatures in various network applications, e.g., credit card authentication by handwriting through digital terminal, increases the needs for the automated writer verification methods. However, there is still room for improvement in the performance conducted by such methods compared to human beings. Since the applications sometimes result in grave consequences, further research to improve the performance of automated writer verification is required. Especially, problems caused by the limited samples have to be solved. In practice, a sufficient number of references per writer are often unavailable. In this paper, we will report the advantage of the user generic model with pen pressure information in such situations. The experimental results which use samples from 54 volunteers show the decrease of error rate from 10.0% to 4.6%.