Intelligent robotic wheelchairs can improve the quality of life who cannot walk independently and handicapped, especially in society of serious aging. In the process of helping people moving from one place to another, how to select the favorite path for users are very important. And existing methods mainly respond to get a safe path for users, which is not enough obviously. To be acceptable to users, the path should not only be safe, comfortable and fast, but also the favorite one for the user. However, it is difficult to identify the favorite path for different users, because everyone will have their preference. Hence, this study proposes a path preference recognitions method based on evidence network for different users. In this paper, human preference was extracted from the user who sitting on the autonomous wheelchair. The user preferences are affected by the width of paths, the number and the shape of obstacles and the distance from the obstacles to users, and they are all quantified respectively. Then all of them are put into network and fused to get the user's preference from complex conditions of different paths. In the end, the results show that there is a fair correlation between the user and the proposed method, and the match rate of the preference for paths is 70.83%.