Research on child care simulation teaching integrating EfficientDet detection algorithm and virtual reality technology

CHEN Tingting; LIU Wen

doi:10.12102/j.issn.1009-6493.2025.24.005

Abstract：Objective:To investigate the effectiveness of child care simulation teaching approach that integrates the EfficientDet detection algorithm with virtual reality technology.Methods:A virtual human-based teaching platform was developed using virtual reality technology.The EfficientDet detection algorithm was enhanced through the incorporation of a spatial attention mechanism and Laplacian pyramid decomposition to enable multi-category action detection.Furthermore,an integrated algorithm combining three-dimensional convolutional neural networks(3D-CNN)with bidirectional long short-term memory(Bi-LSTM)networks was employed to capture spatiotemporal features from video sequences and achieve accurate recognition of pediatric nursing actions.Results:The improved EfficientDet algorithm achieved a mean accuracy of 98.42%,processing up to 25.14 video frames per second.In the task of nursing action recognition,the 3D-CNN-Bi-LSTM model attained average accuracy rate,recall rate,and F1 scores of 95.24%,91.36%,and 93.68%,respectively,outperforming comparative algorithms.Nurses trained using the virtual human teaching platform demonstrated significantly better learning outcomes compared to those receiving traditional teaching.Conclusion:The proposed nursing action detection and recognition algorithms exhibit high precision and reliability.The simulation teaching framework based on these algorithms and virtual reality technology effectively enhance nurses' learning performance in pediatric nursing,offer a novel and promising solution for nursing education.