This paper proposes a behavior analysis system for quantitative evaluation of infants with attention deficit hyperactivity disorder (ADHD) using video images. The system records infantsʼ movements using video cameras without any markers, and quantitatively evaluates each infantʼs behavior using features extracted from video images based on the behavior model. In the experiments performed, behaviors of infants in nursery school were evaluated using the proposed system. The results showed that the behavior of a high-risk infant significantly different from those of normal infants. These outcomes indicated that the proposed system could discriminate the behavior of normal infants and high-risk infants.