Research On Activity Recognition Technology Regardless Of The Phone Orientation And Position

Nowadays,with the widespread use of smart phone,using the accelerometer and gyroscope sensor on smart phone to recognize human activities become a hot topic in the field of pervasive computing.Based smart phones embedded inertial sensor activity recognition,can keep abreast of people's behavior and activity patterns of daily activities,which can provide better personalized service for the people of health care,disease prevention and care of the elderly,improve people's lives and health,Therefore,the activity recognition on smartphone has important theoretical and practical value.In the activity recognition based smartphones,Smartphone orientation and position are two key factors that affect the accuracy of recognition,this article is committed to eliminate the impact of mobile phone orientation changes and mobile phone position changes toward activities recognition.To summarize,the main contributions obtained are as follows:Firstly,for phone orientation change affects activity recognition issue,using the matplotlib tools for sensor data visualization,and in-depth analysis of the impact of phone orientation changes on the sensor data,then presents a dynamic identify three-axis sensor data method to deal with the problem,thereby eliminating the impact of smartphone orientation changes toward human activities recognition.Secondly,for phone position change affects activity recognition issue,proposed a method to recognize the phone position at first and then recognize the activities,which combines the latest smart phone position recognition technology,the characteristic is able to generate a reasonable activity recognition model according to the specific phone position,Thereby reducing the impact of phone position changes to activity recognition.Finally,designed an activity recognition system on the Android smartphone,and then measuring the resource consumption of four kinds of sensor and three activities recognition classifier on smart phone,and then select the appropriate sensor and classifier to conduct a prototype system implementation,The Experimental results show that the system can effectively recognize human activities without limiting the orientation and position of smart phone.