We investigate the QoE-centric heterogeneous network selection problem. Users usually struggle for the maximum throughput when select access network. However, compared to QoS-centric optimization mechanism, in order to get more rational allocation of resources and improve the resource utilization, we seek the QoE optimization method mechanism instead. The core idea of QoE-centric optimization is to allocate network resources according to user needs to make the overall system quality of experience (QoE) improved and reach a stable value at last. In this paper, we first introduce a basic QoE model and then we proposed a QoE game in dynamic environment and analyze its properties. Next, we design a QoE equilibrium learning algorithm. Simulation results show that QoE game framework can effectively improve users QoE and system performance compared with the QoS-centric optimization scheme.