According to the existing problem of breeder continuous observing sows’behavior for sow farrowing judgment, working hard, piglet death due to negligence, monitoring system of sows’behavior before farrowing based on three axis acceleration sensor and WSN was designed. Sows’activities information was collected by using wireless acceleration sensor node. According to different acceleration curve volatility, behavior type was classified by K-means clustering algorithm, displayed real time and wireless transferred by 3G. The experiments showed that the system could acquire and transfer sows’activities information in no time, and the typical behaviors consist of lying, standing, eating and nesting were detected correctly. The accuracy was 87.93%. The system worked stably.