Research On Human Activity Recognition Method Based On WiFi Signal

With the development of wireless local area network technology,wireless sensing based on Wi Fi signals has become a new sensing medium besides vision and sensor,which provides a new intelligent sensing mode for users.It has important application value in the field of intelligent monitoring and human activity recognition.This paper proposes a framework of human activity recognition system based on Wi Fi-CSI signal enhancement.The main research contents of the paper are arranged as follows:Firstly,this paper studies the sensitivity of antennas to human activities.According to the sensitivity of the antenna to the static and dynamic components in the environment,a criterion for judging antenna sensitivity and a dynamic adaptive antenna selection algorithm based on maximum range is proposed.It can automatically select the best antennas from the raw signal,and remove the insensitive antennas data,reducing the interference of insensitive antenna data.Secondly,on the basis of antenna selection,this paper proposes N-iterative signal enhancement and P-order moment enhancement algorithms based on the local variance theory,and compares the performance with the raw signal.The results show that the two signal enhancement algorithms can effectively amplify the active signal,suppress the inactive signal,and increase the difference between the active signal and the inactive signal.Finally,on the basis of signal enhancement,this paper proposes an action interval segmentation algorithm based on interquartile range,extracts the active signal from the raw signal.A human behavior recognition model based on Wi Fi signal is established,and the method proposed in this paper is used for comprehensive experimental verification.Experimental results show that the framework of human activity recognition system can effectively eliminate redundant data caused by insensitive signals and inactive signals,and the overall amount of data that can be reduced by nearly 75%.This paper uses three machine learning methods: random forest,support vector machine,and K nearest neighbors to classify human activity in three datasets.The experimental results show that the average classification and recognition accuracy of N-iterative signal enhancement and P-order moment signal enhancement is 2.8% and 3.2% higher than that of the raw signal,respectively.It has strong anti-interference ability and robustness,and reduces the computational complexity and recognition time.