Finger knuckle print (FKP), known as a biological feature, has drawn great research attention in the field of biometrics recognition. That being said, the development of finger knuckle print recognition is still limited by the lack of data and the difficulties in the extraction of its region of interest (ROI). To resolve these issues, this paper proposes a generative method based on the simulation of the curve distribution of a finger knuckle print to generate reasonable masks of finger knuckle points. Following this, generative adversarial networks (GANs) are applied with the masks to generate the pseudo finger knuckle point images. This method can provide large amounts of training data for recognition as well as directly supplying the region of interest. Experimental results show that the generated finger knuckle print examples can effectively augment the training data for the recognition model.