An AI-based posture and movement recognition system designed at City University of Macau offers a cost-effective way to teach children about movement safety
Falling is one of the most common sources of injury. The impact of falling is often more acute among older adults. For children, falling can be a common occurrence as they develop motor skills and spatial awareness. To researchers at City University of Macau (CityU Macau), this presented an important question: if falls cannot be completely avoided, can we teach people to fall more safely from an early age? The result is a patented system called SafeStepK, which uses AI to teach children how to incorporate safe falls into their muscle memory. The system comprises protective equipment and wearable sensors that capture movement data.
It incorporates OpenPose posture recognition, developed by Carnegie Mellon University in the US, to capture skeletal key points and form a visual movement trajectory of the child’s body as they fall. Using this data, the system analyses contact sequences, impact, head and neck protection, hip loading and rolling paths to provide corrective feedback on posture and movement.
Research suggests that the time between ages six and 12 is crucial for developing motor control, body coordination and movement memory. “If children can learn protective movements such as chin tucking, elbow flexion, lowering the centre of gravity, sliding with momentum and rolling sideways during this stage, these movements may gradually become stable physical memories that support long-term self-protection,” explains Yulin Zhao, assistant professor in CityU Macau’s Faculty of Innovation and Design.
AI is a key component of the system. Its role is not just to determine whether a child has fallen but also to help children understand whether their movement during a fall is safe.
Details such as whether the arms are rigid or the chin is tucked are difficult to notice with the naked eye. AI transforms these movements into actionable feedback. For example, the system can flag the risk of an abnormal impact and angle at the back of the head or neck, reminding the child to tuck in their chin and curl their body.
“AI in SafeStepK does not replace teachers,” Zhao emphasises. “Instead, it serves as an assistive tool for both teachers and children. It turns safety movement instruction – which traditionally depends on experience and observation – into a visualised, data-driven and continuously correctable learning process.”
Training can be personalised according to age, physical development, movement ability and performance data. It starts with low-height and slow-speed movements, working up to continuous movement patterns and progressing to a safe falling movement chain.
By Times Higher Education
