Recently, machine to machine (M2M) services such as smart grid, smart navigation, and health-care are considered as new service models in cellular network. Such an M2M communication causes an instantaneous shortage of uplink capacity due to massive devices and synchronized traffic characteristics. For these M2M properties, cellular network needs to adopt time division duplexing (TDD) system which can adjust the available radio resources according to the uplink and downlink traffic characteristics. However, this TDD system may have different channel condition compared with that of existing systems owing to inter-cell interference scenarios in crossed timeslot; consequently it is impossible to perform scheduling considering actual channel quality. In this paper, we propose an effective scheduling scheme which can improve downlink throughput based on estimation of channel condition of crossed timeslot through cooperation of neighboring evolved NodeBs (eNBs). In the proposed scheme, eNB of downlink cell estimates actual channel condition by sharing scheduling information between neighboring cells and allocates optimal radio resources on the basis of estimated results. By system-level simulation, we show that proposed scheduling scheme can effectively allocate radio resources, and consequently downlink performance can be improved.