Abstract: In order to enhance spectral and energy efficiency in heterogeneous cellular networks, we used distributed stable matching to solve user association and resource allocation problems and utilized a hierarchical sleep mechanism to improve energy efficiency with modeling various base stations by stochastic geometry tools. Firstly, a joint optimization of minimized total power consumption based on the system model was proposed, which was under the constraint of user's QoS request. Then we decoupled it into two sub-problems，which were user association with resource allocation and base station sleep respectively. Finally we used stable matching algorithm and voting-based hierarchical sleep mechanism to solve above problem by oneself. Simulation results show that our scheme can effectively decrease system energy consumption，and has a lower user blocking ratio compared with existing sleep mechanisms.