Independent mobility is an important element for children's development and beneficial in helping them acquire physical and intellectual activities. Unfortunately, existing interface devices for powered wheelchairs are inaccessible to individuals with muscular impairment of their upper/lower limbs or who have involuntary movements of their extremities. This research aims to develop an autonomous mobility device, which adapts the self-learning system to personal characteristic movements. This device incorporates a light sensor to reduce the physical burden and uses a learning algorithm, such as a neural network system. In this paper, we describe the measurement of the user's movement, the establishment of a neural network system, which discriminates the target movement, and the overall system evaluation.