As the potential for a treatment of Duchenne muscular dystrophy (DMD) grows, the need for methods for the early diagnosis of DMD becomes more and more important. Clinical experiences suggest that children with DMD will show some lack of motor ability in the early stage when compared with children at the same age, especially in balance and coordination abilities. Is it possible to quantify the coordination differences between DMD and typically developing (TD) children to achieve the goal of screening for DMD diseases? In this study, we introduced a Local Manifold Structure Mapping approach in phase space and extracted a novel index, relative coupling coefficient (RCC), from gait pattern signals, which were acquired by wearable accelerometers to evaluate the coordination of children with DMD during a walking task. Furthermore, we compared the RCC of 100 children with DMD and 100 TD children in four different age groups and verified the feasibility and reliability of the proposed indices to distinguish children with TD from DMD. T-test results show that, for all age groups, children of the same age with DMD and TD show significant differences in RCC (p < 0.001). Moreover, RCC comprehensively reflects that the coordination ability of DMD patients under walking tasks gradually decreases with age, which is consistent with clinical experience. As a functional biomarker extracted in the phase space of the gait data, the proposed coupling degree index RCC could sensitively distinguish between DMD and TD children at the same age and provide alternative insights and potentially valuable tools for the screening of DMD.