Abstract: How to discriminate the old banknotes from the banknotes is one of important task in financial institution. For the separation of old banknotes, the separation algorithms for the old banknotes was proposed based on statistical feature. Firstly, the training set was formed by the statistical features, which include the standard deviation and the discontinuity degree of the corresponding gray image for banknotes. Secondly, Learning Vector Quantization neural network was used to train the relation between input and output. Experimental results showed that the proposed algorithms can improve the separation accuracy.