With the development of human body perception technology and virtual interactive experience, this technology has also been applied more in the field of driver education. However, the technology is not able to achieve 100 % real simulation in the process of simulating driving movement. In order to solve this problem, the study proposes to improve it by using the body sensing control wash-out filtering algorithm, and at the same time, further optimize it by genetic algorithm, and then carry out a study on the virtual interactive driving education game based on the improved technology. The proposed improved algorithm’s performance comparison results validated that it achieved the highest accuracy rate of 96 %, a final accuracy rate of 92 %, a final loss value of 1.99 %, a running time of 6.5 s, and a mean squared error of 3.8 %, all of which surpass the compared algorithm. After conducting experiments on the actual effectiveness of the designed driving education game, the maximum deviation between the simulated X-direction displacement curve of the game and the actual value was found to be 0.06 m, which outperformed the comparative experimental group. Additionally, both the angular displacement curve and perceived acceleration were the most accurate. In summary, the proposed virtual interactive driving education game based on human perception can more accurately reproduce the real scene during the driving process, providing users with better learning conditions.