关键词: 数字信号处理, 图像处理, 特征提取, 人工神经网络

Abstract: A banknote classification system is designed and implemented for US dollar system.The classification system can process reliably the US dollars in high speed.Banknote sorter is one kind of the automatic financial tool.Banknote image analysis technique is hard-core of the sorter.It is a complex photic & mechatronic system.Its hardware is composed of image sensor，signal capture controller and digital signal processor.After capturing the banknote image with high speed，the DSP completes the image location，declining adjusting，feature extraction and image analysis.The banknote classification is finished by BP neural networks.The system obtains good results in the practical application.It can processes the banknote at speed of 800/min without misclassification.

Key words: Digital Singnal Processor（DSP）, image processing, feature extraction, neural networks