Spatio-Temporal Transformer with Hypergraph in Nursing Activity Recognition

In today's aging population, the safety of elderly patients is a concern for society, which increases the demand for nurse care. By recognizing nursing activities, we may not only relieve nurses of the burden of maintaining manual records, but also provide more advanced educational methods for the field of nurse training, as well as technical support for the development of nursing robots in the future. In this work, we present a hypergraph transformer-based network, which extracts features from skeletal joints to perform nursing action recognition. Under the condition of solely utilizing skeletal data, our approach achieves state-of-the-art accuracy of 79.3% on the Nurse Care Activity Recognition Challenge dataset.