The effective body impact detection of quadruped robot can support the dynamic balance control decision, but related studies have not been appeared in the literature. A novel body impact detection algorithm for quadruped robot is proposed. The leg impact acceleration vector and gravity acceleration vector which expressed in the body coordinate were obtained by dynamic analysis and attitude calculation which based on Kalman filter, and both of them were separated from accelerometer output signal to get the body impact acceleration. Finally, the dynamic simulation is conducted to the quadruped robot, and the result verifies the validity of the proposed algorithm.