With the development of biosensors and wireless technology, the early wearable medical devices have been far from meeting the demands of human, and the mobile medical care has recently gained the increasing attention. The wireless body sensor network (WBSN) has made a significant contribution to the non-invasive collection of human physiological signals. This paper aims at the collection of four vital signs of the human body in WBSN, different priorities are given to four vital signs and the queuing model of different priority data that collected at patient's wireless device (named PDA) is analyzed. Meanwhile, this paper notices that the problem of network congestion in WBSN, the network environment of data transmission is divided by the learning results from mote using learning automata. In order to control network congestion, dropping packets of PDA is guided by mote under different network environment and the system delay after dropping packets at PDA is analyzed. Results in this paper show that “learning” at mote can significantly reduce system delay.