Abstract: Adapting to the limited resource of mobile device, a human state recognition algorithm based on Kalman filter was proposed, which could identify dynamic, static and state transition in real time. The Bluetooth module with a triaxial accelerometer was placed on the chest of body to collect three-dimensional acceleration data. The characteristic of human activity was associated with the features of the accelerometer signal, so the function of change of the signal vector magnitude (SVM) was processed by Kalman filter to identify human state. Experiment results show that the algorithm achieves high accuracy in identification of postural transition, meanwhile, the algorithm has displayed better performance with little overhead on the smartphone.