Research On Detection And Recognition Of Indoor Multiplayer Activity Based On WiFi

With the rapid development of artificial intelligence,human-computer interaction technology has attracted much attention.In order to make it easier for computers to understand people's intentions and achieve more convenient and intelligent interactions between people and computers,the detection and recognition of human activities and behaviors is particularly important.However,traditional human activity recognition technologies,including wearable devices and vision-based recognition methods,have some limitations and disadvantages.They have a certain degree of intrusiveness,and they also require higher costs.Nowadays,wireless network technology has developed rapidly,and wireless WiFi devices have become popular.They are installed in many indoor locations such as homes and offices.They are low cost,easy to deploy,and widely distributed.They are one of the largest wireless sensor networks.In this paper,combining its advantages,a method of detecting and recognizing indoor human body using WiFi signals is proposed.By collecting CSI data of WiFi wireless signals,and processing and analyzing these data,a machine learning classification algorithm is used to detect and identify indoor human activities.The main work of this article is as follows:(1)A method of detecting and identifying indoor human activities using CSI data of a WiFi signal is proposed,and a system architecture model is designed.Compared with the traditional identification methods,this method is passive detection and identification,non-invasive to the human body,and low cost,easy to deploy.(2)Use WiFi devices to collect CSI data,extract both amplitude and phase signal information,and perform denoising and phase correction processing.Compared to using only the amplitude,the combination of amplitude and phase can capture more information on changes in the WiFi signal and can capture more fine-grained human activity.(3)An anomaly detection algorithm ADA for indoor human activities was designed.According to the characteristics of CSI data changes,by analyzing the subcarrier fluctuation characteristics of the received signal,this method can more accurately distinguish the static state(no activity)and the human activity state,and on the basis of this algorithm,proposes a detection method to exact the start and end time points of human activity.Moreover,the detection of activities can adaptively regulate different levels of sensitivity to adapt to different surroundings.Experimental results show that the detection algorithm in this paper can achieve an average activity detection accuracy of 97%.(4)Classifying and identifying single person activity and multiple activity combinations of multiple persons respectively.A speed model for activity recognition was proposed to quantifies the correlation between CSI value and the movement speeds of different human activity.Obtain the amplitude and phase information of the 30 sub-carriers of the WiFi signal.and perform active feature data processing,and then use the kNN machine learning algorithm to classify and identify.The experimental results show that for a single person,the three categories of activities can achieve an average recognition accuracy of 95%,and for an average of nine combinations of multiple activities,the recognition accuracy rate is 91%.