To investigate the performance of different neural network algorithms in identifying serial number of RMB banknote, the training speed, recognizing speed and rate, and ability of anti-noise of five neural network algorithms, including the discrete Hopfield neural network, BP neural network, PNN neural network, GRNN neural network and SVM neural network, are studied. The simulation results show that amongst the five algorithms, BP performs worst, followed by SVM and Hopfield, with PNN and GRNN performs best, not only gives the higher recognition rate, shorter training and recognition time, but also is more robust to noise.