Six-axis acceleration sensing mechanism is a kind of inertial sensors getting kinetic parameters in the field of aerospace, robots and the like, and researches on it at home and abroad are still at exploring stage. Based on the Kane dynamics method, the dynamic equation of the system is established, and the problem is transformed into two initial value problems of first-order ordinary differential equations, which are solved by the improved Euler method. Based on local minima and feature threshold of observed quantities, the criterion of feature points is given, and then an error auto-compensation algorithm of semi-closed-loop structure is established. Simulation results show that the maximum errors of the linear acceleration and angular acceleration are 0.0569% and 0.0054%, respectively, thereby demonstrating the effectiveness of the proposed decoupling algorithm.