In this paper, we propose a new approach for real-time interactions with 3d model using a data glove and a position tracker. Recently, hand gestures are considered as one of the most meaningful and expressive signals. Since most of the vision-based hand gesture recognition methods are sensitive to the environment and require specific light condition, we choose data glove to interpret the hand gestures. Using the glove-based gesture recognition method, accurate and real-time interactive performance can be obtained. Additionally, the collision detection algorithm is introduced to enhance the object manipulating tasks. By using the gesture and collision detection based methods, several manipulation modes are designed and verified, such as translating and rotating a model, creating a clipping plane, changing the viewpoint, and grasping and disassembling a model group. The results show that the glove-based interactions are more natural and accurate than the vision-based hand gesture methods.