面向人体运动监测的无线体域网中信号压缩与动作识别研究

Wireless body sensor network(WBSN) has recently gained tremendous interest among a range of health-related areas. Energy efficiency is an important challenge to the devlopment of WBSN for continuous long time of physical health monitoring. In this project, taking the human movement monitoring as an application, signal compression and activity recognition are taken into consideration as a closely related whole by combing the sparse representation and compressed sensing. The project aims to prolong the life of WBSN under the constraints of limited sensor node calculation, storage resources and wireless body area network bandwidth while meeting physical activity monitoring requirements at the same time. Research contains the following specific lines of investigation: (1)explore the inner link between signal compression and activity recognition, construct the correlation model. (2)explore on a novel joint sparse representation method for activity classification of multi-sensor, which takes into account the spatial and temporal correlations between sensors. (3)investigate the structure of the motion signal sparse characteristic, construct an adaptive coordation mechanism of the activity recognition and compression perception. The proposed research will help not only on energy efficient information perception and calculation but also on the application of sparse representation and compressed sensing. The project outcomes will have broad application values such as in the areas of the remote medical care and the monitoring for special population.

无线体域网作为一种新的普适医疗保健、疾病监控和预防的解决方案成为研究热点。要实现长时间不间断监测，能量有效性是体域网面向健康监测应用所亟待解决的问题。本项目以人体运动状态监测为应用切入点，融合压缩感知和稀疏表示理论，将传感信号压缩与动作识别作为密切相关的整体联合考虑。在满足身体活动监测要求的前提下，实现在传感器节点计算、存储资源和无线体域网带宽有限的约束下，达到节能的目的。具体研究包括：研究压缩的传感数据与动作识别之间的内在关系，构建数据压缩与动作识别的关联模型；研究多传感器采集的运动数据之间的时空相关性，构建基于多传感器的分布式压缩感知和多运动信号的联合稀疏表示的动作识别机制；研究运动信号的结构稀疏特性，构建动作识别与自适应压缩感知的联合框架。本项目为无线体域网中能量有效的信息感知与计算提供新的思路，在远程医疗保健、特殊人群监护等领域有广泛的应用价值。

能量有效性是无线体域网在人体健康监测应用中面临的重要挑战。本项目以无线体域网在人体运动状态监测中的应用为切入点，围绕着能量有效的运动信号压缩和识别问题展开研究。本项目借助压缩感知和稀疏表示理论，将传感信号压缩与动作识别作为密切相关的整体联合考虑，实现了在满足身体运动监测要求的前提下达到节能的目的。具体研究成果包括：1）针对运动信号的稀疏性，对压缩的数据建立稀疏表示的人体运动模式识别模型，提出了一种压缩分类的动作识别方法。2）针对多传感器采集的运动数据之间的时空相关性，构建了基于多传感器的分布式压缩感知模型和多运动信号的联合稀疏表示模型，采用层次贝叶斯模型来求解稀疏表示系数来实现更高的动作识别准确率和更大的数据压缩率。3）针对运动信号的稀疏度的时变性，提出了基于动作识别结果反馈的自适应压缩感知框架，建立了衡量能量和识别率的最优化函数来进行识别准确率与能量有效性的权衡，自适应地确定最优投影数量。4）设计并实现了用于人体运动监测的无线体域网原型系统，采集了人体日常活动中的动作数据，对传感器节点的能耗进行了测试与分析，实际验证了所提出的动作识别算法的识别准确率和能量有效性。5）项目发表SCI期刊论文3篇，国际学术会议论文1篇，培养博士研究生1名，硕士研究生4名，申请国家专利1项，基本完成项目的预期目标。本项目的研究成果对无线体域网中能量有效的信息感知与计算具有参考价值，在远程医疗保健、特殊人群监护等领域具有应用价值。