Research On Activity Recognition Method Based On Indoor Location Service

In recent years,with the rapid development of Human-Computer Interaction(HCI)technology,wearable sensor products are continually emerging.Its small size,easy to carry,and strong computing power make it a key application in fall monitoring,elderly care,and office behavior monitoring.At the same time,modern people spend about 70% of their time indoors,and indoor location services have grown up to become a basic requirement.If the human activity and state information can be obtained through technological means,and human intent under the activity can be judged.The user can be made available to personalized service.As positioning accuracy will gradually exceed the centimeter level,positioning technology will bring greater benefits to the research of behavior recognition.Therefore,research on activity recognition based on location service sensors has become a research hotspot.However,in the current position sensor based on activity recognition research,due to the low positioning accuracy,the wearing method without any standards and limitations,and the use of different,resulting it in low accuracy of activity recognition,single activity and poor universality.In summary,this paper will develop from the activity recognition method and the sensor layout scheme.The purpose is to get the advantage of the three-dimensional coordinates of the body part in real time by means of the position sensor to enhance the recognition ability of complex activity and the universality of the activity recognition method.Firstly,aiming at the problem of low recognition accuracy and single activity identification in the activity recognition research using position sensors,this paper proposes a layered activity recognition method based on position sensors.By analyzing the characteristics of the human motion model,the human motion model is divided into three layers.Based on the general activity recognition framework,a hierarchical activity recognition method with three layers is designed and identified.The method identifies the basic walking,sitting,lying,and standing state of the human body by the sensor tag feature worn on the trunk layer,and finally determines 16 kinds of daily activities for identifying the human body according to the sensor tag characteristics of the upper and lower limb layers and the relative relationship between the tags.The method effectively improves the accuracy of activity recognition and can identify more complex daily activities in the room.Secondly,a sensor layout optimization method based on quantum genetic algorithm is proposed for the insufficiency of recognition accuracy and poor universality caused by the wearing unclear method of sensors and the diverse uses of activity recognition.Analyze the existing activity recognition requirements,and analyze the relationship between UWB-Tag labels wearing position and number and each requirement for the common three requirements,and use quantum genetic algorithm to find a suitable sensor optimization wearing schemes.Under the obtained wearing scheme,the requirements of universality of activity recognition and recognition accuracy are satisfied.Finally,the layered activity recognition method based on the position sensor and the sensor optimization method based on quantum genetic algorithm is verified in the experimental scenario.The experimental results demonstrate that the proposed method is feasible.Applying the proposed two methods to the actual scene,compared with the traditional activity recognition method,the proposed method has improved by about 4%.Which is a good solution to the disadvantage of low recognition rate and universality in the identification research using position sensors.