With the increasing of computing and network technology, there has been a rise in children's educational and entertainment software. However, these software applications depend on mouse and keyboard input, which may be tedious and challenging for children. Moreover, these applications fail to capture the interaction between users and computers. This study proposes design principles for children's human-computer interaction based on children's cognitive, psychological, and physiological attributes, and outlines a framework accordingly. The framework integrates natural human-computer interaction technologies, such as physical operation interface, gesture interaction, low-level semantics-based human-computer interaction, and high-level semantics-based human-computer interaction, with the aim of facilitating seamless and intuitive interaction between children and computers. Finally, the results demonstrating the feasibility and effectiveness of the proposed human-computer interaction technology as a child-centered virtual reality application tool.