Computer Engineering and Applications ›› 2017, Vol. 53 ›› Issue (4): 176-183.DOI: 10.3778/j.issn.1002-8331.1507-0126

Previous Articles     Next Articles

Strong robustness human activity recognition based on wearable sensors

LIU Jinyi1, ZHANG Le1,2, HU Haibo1,2, ZU He3   

  1. 1.School of Software Engineering, Chongqing University, Chongqing 401331, China
    2.Key Laboratory of Dependable Service Computing in Cyber Physical Society Ministry of Education, Chongqing 400044, China
    3.School of Telecommunication Engineering, Chongqing University, Chongqing 400044, China
  • Online:2017-02-15 Published:2017-05-11


刘锦怡1,张  乐1,2,胡海波1,2,朱  贺3   

  1. 1.重庆大学 软件学院,重庆 401331
    2.信息物理社会-可信服务计算 教育部重点实验室,重庆 400044
    3.重庆大学 通信工程学院,重庆 400044

Abstract: Human daily activity recognition using mobile personal sensing technology plays a central role in the field of pervasive healthcare. In this paper, a novel human activity recognition framework is presented to reduce the impact of sensors displacement. It utilizes high-precision sensor to capture signal. According to periodic features of human movement, the corresponding frequency domain is got by Fast Fourier Transform. The principal components analysis is used to extract composite indicator. After extend process the input data, a self-organizing neural network models is built for gesture recognition. Experimental results demonstrate the effectiveness of the scheme, and in ideal conditionsthe accuracy of certain relationship can get 97.5%.

Key words: fast Fourier transform, self-organization neural network, principal component analysis, human activity recognition

摘要: 为了降低可穿戴传感器在传感器移位时对动作识别率的影响,对可穿戴传感器的动作识别进行了研究。采用高精度传感器采集不同部位的输出信号,根据运动的周期特点对输出信号进行去噪和快速傅里叶变换,将其转化为频域信号。再使用主成分分析法提取综合指标,并对自组织神经网络进行训练,实现动作识别。最差情况下识别准确率可达到92.0%,较好情况下甚至可达到97.5%,传感器移位情况下的识别率甚至更高。

关键词: 快速傅里叶变换, 自组织神经网络, 主成分分析, 动作识别