Since there are severe instability problems of human-computer interaction in 3D scenes' navigation, a data processing algorithm of variable smoothing and normalization model for sensors of smart terminals was proposed. In view of the data characteristics of sensors and processing characteristics of each smoothing model, the different smoothing models were combined and the variations of smooth window were enabled. A normalization algorithm was applied to deal with the interaction data of sensors equidistantly. The experimental results in navigation system of 3D scenes based on terminals show that, the standard variance decreases by 71.3% after dealing with low frequency data of sensors with the proposed algorithm, which means the data perturbations is significantly reduced. And while processing the high frequency data, the standard variance decreases by 7.9%, which effectively preserves the signal features. The proposed algorithm can distinctly improve the stability and continuity of 3D navigation interaction using smart terminals.