基于改进人工神经网络的人体姿态识别方法在人机交互医疗设备中的应用

为提升人机交互医疗设备对久坐不动、常年卧床等状态下人体的监测效果,在利用无线体域网(Wireless Body Area Network, WBAN)建立人体姿态识别系统的基础上,设计了相应的改进人工神经网络与WBAN系统进行融合,并将其应用于人机交互医疗设备中。结果表明,在HiEve数据集中,该方法于20次迭代时开始收敛,损失函数值为0.0112。在患者不同姿势的识别验证中,该方法下的人机交互医疗设备识别准确率均显著高于90%,并且耗时最短仅为23.16s,具有较高的识别准确率和效率,为人体姿态识别及相关医疗设备的应用提供了更为可靠的技术参考。

Application of improved artificial neural network-based human pose recognition method in human-machine interactive medical devices

In order to improve the monitoring effect of human-computer interactive medical equipment on human body in the condition of sedentary and bedridden all the year round, based on the establishment of human body posture recognition system using wireless Body area network (WBAN), a corresponding improved artificial neural network is designed to fuse with WBAN system, and is applied to human-computer interactive medical equipment. The results show that in the HiEve dataset, the method starts to converge at 20 iterations, and the Loss function value is 0.0112. In the recognition verification of different patient postures, the accuracy of human-machine interaction medical device recognition under this method is significantly higher than 90%, and the shortest time is only 23.16 seconds. It has high recognition accuracy and efficiency, providing a more reliable technical reference for human posture recognition and related medical device applications.