Realizing facial expression is likely to be one of the earliest facilitators of social engagement, and it is a core deficit for individuals with autism spectrum disorder (ASD) to build social relationship. Mobile learning creates the possibility of a learning environment and visual stimuli that can adapt its informal learning contexts to improve recognizing emotions for people with ASD. This study adopted features of mobile learning to help the realization of facial expression for the ASD. The proposed 3D Complex Facial Expression Recognition (3CFER) system was developed to help the deficit in facial expressions for this population. This study therefore, explored how children with ASD performed realizing facial expression using the 3CFER system; and how the phenomena of using the learning system for people with ASD performs. Participants (n = 24, 16 males and 8 females, m = 11.3 years old) were randomly assigned to either a control or an experimental group, and were involved with the pre-and post-test sessions. The control was not engaged in the system-treatment; and the experimental undertook the system-operation. The result showed that the experimental had great improvement in realizing facial expression compared with control, and "surprise" and "shyness" were mostly easy to be identified for them. The performance of using mobile learning system was promising well. However, the informal experience of recognizing facial expression in different social contexts was meaningful learning for this population.