Design Of A Wearable Human Action Recognition System With Dynamic Time Warping

In the time of big data,the motion recognition technology is widely used in competitive sports,health testing and medical research.The traditional motion pattern recognition technology is mostly based on vision.but the expensive equipment has become an important factor limiting the development.With the development of the Micro-Electro-Mechanical System(MEMS),the price of inertial sensors is getting lower and lower,while the accuracy has been improved.Therefore,the motion pattern recognition based on inertial sensors becomes more and more popular.In this paper the motion pattern recognition based on wearable sensor is studied,and the dynamic time warping(DTW)algorithm is used to recognize and analysis the motion of people.The research content is divided into the following aspects:Firstly,we design a set of wearable motion information acquisition system,using accelerometer?gyroscope?magnetometer and the pressure sensors to get the motion data.Then we use the computer to collect the motion information from the system.Then by analyzing the human skeleton model,we describe the human kinematics based on quaternion and rotation matrix.The daily movements of us are analyzed and expressed by the feature signals.It's the basic of the pattern recognition algorithm.Finally a motion pattern recognition algorithm which is based on the dynamic time warping is proposed.Firstly it recognizes the basic feature state by using the decision tree.Then use these states to segment the motion data.Finally use the dynamic time warping to realize the motion pattern recognition.The results of experiments show that this method has a good recognition effect.