Open AccessHuman Motion Recognition Based On Inertial Sensor
Author(s) -
Mengmeng Xing,
Guohui Wei,
Hui Cao,
Feng Yang,
Yongqi Nie,
Jing Liu
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/782/3/032099
Subject(s) - support vector machine , artificial intelligence , inertial measurement unit , activity recognition , computer science , pattern recognition (psychology) , entropy (arrow of time) , inertial frame of reference , principal component analysis , human motion , extreme learning machine , machine learning , motion (physics) , computer vision , artificial neural network , physics , quantum mechanics
Human activity recognition (HAR) has attracted considerable research attentions from all walks of life due to its application in human-computer interaction, smart medical treatment and smart home health care. There are various ways of using different motion capture methods for HAR. Among which, HAR based on inertial sensor signals has recently emerged as a challenging and hot research topic. In this paper, we presented a inertial sensor-based method for HAR. Firstly, the variance, mean, information entropy, peak, etc. were extracted from the inertial data as robust features of the actions. Then, the robust features were further processed by Principal Component Analysis (PCA) to reduce the dimensions of features. Finally, the features with different actions label were used to train the Extreme Learning Machine (ELM) and Support Vector Machine (SVM) model. The result can achieve an mean accuracy of 83.49%, it also verified the effectiveness of our method.