Research On Human Body Behavior Recognition Algorithm Based On Mobile Device Motion Sensor

Wearable computing which is a new computing model of Human-Computer Interaction,has been widely used in many fields such as military,public health,electronic consumption,sports,education and so on.Key technologies of wearable computing have become a hotspot in academia and industry.Human behavior recognition technology is an important research branch of wearable computing.In this thesis,creative works are carried out on the motion acceleration sensor based human body behavior recognition algorithm.Data acquisition,data preprocessing,machine learning classification model and behavior recognition algorithm are studied.The main contents are as follows:(1)This paper explores the use of smart phones that has a built-in motion sensor to collect human behavior data.The obtained three-axis acceleration sensor data should be preprocessed in order to remove the noise,and be split into data segmentations.(2)The characteristics in time domain and frequency domain of typical human behavior,which include walking,upstairs,downstairs,sitting,standing and falling,are extracted respectively.Arbitrary pairs characteristics in time domain and frequency domain of these typical human behavior is comparative analyzed,the deeper detail of characteristic distinction help to establish the behavior recognition characteristics data set;(3)Inspired by the traditional human behavior recognition algorithm based on support vector machine model,the support vector classifier and the second kernel function theory are combined to construct the quadratic kernel support vector classifier model through theoretical analysis.The quadratic kernel support vector classifier model based human behavior recognition algorithm is proposed,and simulation results show that the proposed algorithm has more accurate recognition rate than the existing human body behavior recognition algorithm based on random forest classification model.