【摘要】 基于传感器的行为识别是环境智能中的一个关键问题,但是"该使用哪些种类的传感器?"以及"每种传感器该使用多少个?"都还是有待解决且具有挑选性的问题。针对这些问题,提出一种基于特征选取的方法来选择传感器。实验结果显示,在OPPORTUNITY标准数据集上,1)相比于加速度传感器和陀螺仪传感器,磁性传感器的识别效果较好;2)只使用被选取的5个最好的磁性传感器特征也能达到和使用全部传感器特征类似的识别准确度。更多还原

【Abstract】 Activity recognition using sensors is a key issue for ambient intelligence. However,what kinds of sensors we should use and how many of each we should use are still challenging problems to be solved. In light of these,in this paper we propose to use a feature selection-based method to select sensors. Experimental results demonstrate that,on OPPORTUNITY benchmark dataset,1) the performance based on magnetic sensors is better than that based on acceleration sensors and gyroscope sensors; and 2) it can achieve similar recognition accuracy by using only five best magnetic sensor features instead of using all sensor features.更多还原