基于传感器数据的人类活动识别研究

Review of human activity recognition using sensor data

给出了活动识别的定义,阐述了活动识别对智能健康监护和老年人护理的重要意义,并给出了基于传感器的活动识别系统的构成,详细描述了系统的传感器数据采集、数据预处理及机器学习等技术的研究进展,介绍了人类活动识别实验取得的结果。最后指出,老年人活动识别、多人活动识别以及实时活动识别是未来活动识别的发展方向;活动识别研究已成为普适计算一个重要和富有挑战性的研究课题,构建用于处理复杂的现实情况和环境的可靠的活动识别系统仍然是一个挑战,需要多学科交叉研究。

The definition of activity recognition is given, and the significallce of activity recognition to intelligent health- care and elderly healthcare is interpreted. The main parts of an activity recognition system based on sensors are de- scribed, and the corresponding techniques of sensor data sampling, data preprocessing and machine learning, as well as their developments, are reviewed in detail. The review points out that elderly activity recognition, collective activity recognition and real-time activity recognition are the possible directions of future activity recognition re- search, the research on activity recognition is an important and challenging research topic in pervasive computing, and building reliable activity recognition systems to deal with complex real-life situations and environments is still a challenge, which requires a multi-disciplinary effort.