The present complicated heterogeneous network environment is resulted from coexisting of many radio access networks. A very important issue is how to select mutually for users and networks under this condition to maximize each profit. The existed researches mainly analyzed in two angles, one is how to select an efficient and suitable network to access from the point of view of users, the other is how to select users to maximize the profit of the networks from the point of view of networks. However the analysis from either of those angles do not consider enough about mutual selection between users and networks, which may lead to sacrifice one's profit to get another one's optimization. A game theoretical approach of the network selection algorithm is proposed, in which the different requirements and targets of users and networks are considered synthetically. The approach guarantees the profit of both sides through obtaining stable matching by a matching game between users and networks. The simulation results show that the algorithm improves the satisfaction degree of networks/users compared with the algorithms based on user/network decision. Furthermore, the trade-off between the satisfaction degree of users and networks can be reached.