The Research On Activity Recognition Technology Based On Multi-source Data And Logical Reasoning

Today,a new round of scientific and technological revolution and industrial changes are booming,and we have gradually transitioned from the Internet era to the Internet of Things era.At the same time,with the development of sensor technology and the improvement of mobile computing capabilities,smart terminals,as a key component of the Internet of Things,are equipped with a variety of mobile smart devices and entered people's lives.The use of these devices to identify and record human daily behaviors plays a vital role in health testing,assistance for people with disabilities,and care for the elderly.Existing work has various deficiencies.For example,behavior recognition technology that uses radio frequency signals as a sensing medium is very demanding for application scenarios,behavior recognition using machine learning technology takes up too much computing resources,and behavior recognition using situational awareness requires too many sensors or additional devices.In view of this,this paper starts a research on behavior recognition technology based on multimodal data and logical reasoning.Unlike other previous studies,this technology has designed a lightweight perception system that realizes behavior recognition through logical relationship judgment.Combine logical reasoning with body area network to accurately depict the basic behaviors of users in daily life.Innovative use of signal processing technology and logical reasoning to determine the logical relationship between actions.Therefore,a large amount of data pre-training is not required,which reduces the calculation cost and improves the scalability of the system.At the same time,this paper designs an adaptive algorithm of logical relationship,which can better resist the influence of the difference of users' posture or habits on the recognition results.Recognition can also be made accurately in real time among people with large differences.We implemented a system based on this technology using commercial smartphones,smart watches,smart glasses,and a Bluetooth module with an IMU,and did a lot of experiments in real-life scenarios to evaluate performance and verified the reliability of logical reasoning for behavior recognition.We have identified a total of 12 movements of the upper and lower limbs,and the experimental results show that the recognition accuracy rate before and after adding the adaptive algorithm has improved by an average of 7.4%.At the same time,the accuracy of recognition in three different scenarios is also compared,which are 92.3%,93.7% and 92.9%.The results show that the recognition effects of the three are similar.The system in this paper can resist the interference of different environments on behavior recognition and break the barriers of application scenarios.