Research On Method Of Human Behavior Monitoring Based On Wearable Devices

Human behavior monitoring based on wearable devices is a process of extracting the movement status of users by perceiving human activity data with wearable devices and computer technology.In the research of human behavior monitoring methods,the main techniques are data acquisition and behavior recognition.As for data acquisition,most traditional behavior monitoring methods are based on video images.However,this method is limited by many restrictions of situation.Besides,it is intrusive,which can not meet the needs of behavior monitoring.Wearable devices have achieved good results in data acquisition technology.However,most of the researches involve multiple sensors,which do not fully consider user experience.As for behavior recognition,traditional methods can only recognize human activities offline.While under the present condition,real-time monitoring results are required all kinds of applications.In view of the above two problems,we discuss the problem of real-time behavior recognition efficiently using a single sensor.Firstly,we give the method of data preprocessing according to the data collected by a single three-axis accelerometer.Aiming at the noise data which does not overlap with the signal,we first use Butterworth filter to remove the data.Then,we propose a KGA algorithm for removing abnormal data from acceleration and smoothing data at the same time.This method uses genetic algorithm to encode the parameters of Kalman filter to optimize it.Secondly,we propose a human behavior recognition method based on improved One-Dimensional Convolutional Neural Networks(1D-CNNs).According to the motion characteristics,we extract the eigenvalues which can distinguish the types of activities.At the same time,we propose a sample autonomous learning method,which aims to find the optimal sample training set and avoid over-fitting problems in traditional CNNs.In the recognition of 11 human activities,the average accuracy is98.7%.Compared with other behavior recognition methods in the same dataset,better classification is achieved by this method.Thirdly,we propose a threshold-based method for identifying falls that are harmful to the elderly.The key point of this method is to distinguish falls from people's daily activities.According to the characteristics of human falls,we extract eigenvalues that can effectively distinguish daily activities from falls.In addition,we use cross validation to determine the threshold of the method.The results show that inthe analysis of 11 kinds of human daily activities and 15 types of falls,our method can distinguish 15 types of falls.The recognition recall rate in our method reached99.1%?Finally,we design and implement a human behavior monitoring system based on our research.This system mainly verifies the proposed method in real time.The system first uploads the human activity data collected by sensors to the cloud platform,and then the background servers analyze and process the data.Finally,the system identifies the user's ongoing activities and alarms when the user falls.The platform is achieved a recognition recall of 89.97% in our experiments,which demonstrates the effectiveness and efficiency of our method.