In order to achieve efficient communication in the fifth generation (5G) networks, non-orthogonal multiple access (NOMA) technique has been utilized in fog radio access networks (F-RANs). In this paper, we investigate the resource allocation problem in a NOMA-enabled downlink F-RAN. To maximize the weighted sum rate of NOMA users served by fog-computing-based access points (F-APs), the resource block (RB) allocation and power allocation are optimized. Specifically, we decouple the problem into RB allocation and power allocation problems. The former is modeled as a many-to-one matching game and we propose a modified swap-enabled matching algorithm to solve it, which takes interference threshold into consideration. The later is a non-convex problem, we transform it into a tractable one via some approximations and get the closed-form expressions of power allocation coefficients. Finally, we combine the both to propose a joint resource allocation algorithm, which is preformed iteratively to obtain the optimal result. Simulation results are provided to show the performance of the algorithm.