Abstract: Sparse Code Multiple Access (SCMA) is one of the up-link (UP) air-interface technologies for wireless. The Message Passing Algorithm (MPA) is the main method of multiuser detection in SCMA systems. The MPA updates the all codeword messages with iteration. After probabilities of all messages have converged, the iteration is over. So the MPA has a high degree of complexity when all messages are updated in each iteration. In order to solve this problem, an scheme of multiuser detection for SCMA based on dynamically selecting message to update is proposed in this paper by using the characteristic of the probability of each codeword message has different convergence rate. Codeword messages with fast rate were found in each iteration and would not be updated in later iterations because they were close to convergence, other codeword messages continued to update. So the complexity of proposed scheme was decreased. Simulation results showed that compared with the MPA, the Bit Error Rate (BER) performance of the proposed scheme was almost the same as that of the MPA, and the complexity of the algorithm was significantly reduced when the appropriate gravity factor was chosen. Therefore, the proposed scheme has obvious advantage in the application of SCMA systems when the hardware complexity is reduced and performance is maintained.