Currency is an indispensable part of our daily life. However, how to identify true and fake currencies has become the most important issue at present. If we use a computer for currency recognition, it will greatly improve the accuracy of recognition and reduce people's workload effectively. In recent years, deep learning has become the most popular research direction. It mainly trains a dataset through deep neural networks. There are many different models that can be used in this research project. Throughout these models, the accuracy of currency recognition can be improved. Obviously, such research methods are in line with our expectations. In this thesis, we mainly use Single Shot MultiBox Detector (SSD) model based on deep learning as the framework, employ Convolutional Neural Network (CNN) model to extract the features of paper currency, so that we can much accurately recognize the denomination of the currency, both front and back. Our main contributions are: (1) through using CNN and SSD, the average accuracy of currency recognition is up to 96.6%; (2) in order to ensure the recognition, we selected two models for comparisons. One is MobileNet and the other is faster R-CNN. However, we found from the experimental results that, in general, CNN is much suitable for our currency identification requirements. When a currency is tilted or moved, its denomination and front/back side can still be identified.