This article combines traditional scheduling algorithms' characteristic with newly multiplexing method for stream media, integrates two kinds of multiplexing models with monopolizing buffer and sharing buffer, and an efficient statistical multiplexing strategy is proposed that is suitable for VBR real-time stream. This multiplexing strategy creates different weight formulas for two buffer levels respectively: in the first level non-loss multiplexing is carried on, and weight factor is calculated by data storage and arrival order from each channel, so who has the largest weight value will be dispatched at top priority; the second level is compatible non-loss with loss multiplexing, and weight equation is established by the importance as well as overflow of data, finally who has the smallest weight value will be dropped firstly. Through theoretical analysis and simulation test, it indicates that this kind of statistical multiplexing strategy which uses least buffer and does less harm to data can carry the maximum transport rate, thus is more advantageous in raising channel utilization.