Intelligent Nursing Bed for Autonomous Care Based on Low-Cost Resource-Constrained Microcontroller With On-Device Learning

摘要

Due to the escalating worldwide aging problem, nursing beds (NBs), which are crucial amenities for supporting the everyday activities of the elderly and disabled, are being utilized in an expanding range of environments. To alleviate the nursing pressure on caregivers and maintain a positive health mindset for the users of NBs, enhance their self-esteem and confidence, article presents the design of an intelligent NB based on a low-cost resource-constrained microcontroller that enables users to achieve self-care. As the aggregation of sensor usage, the intelligent NB has been specifically designed with two man-machine interaction methods that are beneficial for self-care: speech recognition based on sound sensor and hand gesture recognition based on visual sensor. To perform hand gesture recognition inference on a microcontroller that has limited resources and is inexpensive, we developed a convolutional neural network-neural circuit policies (CNN-NCP) model specifically designed for on-device learning. This model has a small footprint, measuring just tens of kilobytes. Additionally, we created a special dataset consisting of hand gesture images to train and test the model. Experimental results demonstrate that the model achieves a hand gesture recognition accuracy of up to 96.65#x0025;. In addition, this article also discusses the selection, circuit design, and performance testing of the detection and monitoring sensors used in the NB. The realization of all functions of intelligent NB for autonomous care is based on a resource-constrained microcontroller, representing a beneficial and highly challenging endeavor in pursuit of low cost and high reliability.