As the advance of technologies, the smart phone makes the context-aware more easily. And context awareness is an important factor to be considered for content personalized recommendation. Most traditional recommendation systems only take the user preferences into account, which leads to low content prediction recommendation accuracy. In this paper, we conceive a context-aware media content personalized recommendation scheme for community networks, taking users' profiles, behaviors, network conditions and contextual information into account. We classify the context information with the naive Bayesian network. Specially, we design an efficient BP neural network training and predicting scheme for content rating. Good performances of the proposed scheme are verified through a series of experiments. Simulation results indicate that the proposed scheme achieves the improved accuracy of recommendation notably and as well a better user experience in contrast to traditional recommendations.