Research on CSI Indoor Personnel Behavior Detection Algorithm Based on Adaptive Kalman Filter Yanxing Liu a, * , Shuyang Hou b , Xiaoqin Li c and Longyu Shi d College of Computer Science and Engineering, Northwest Normal University, Lanzhou 730070, China a lyanxing@nwnu.edu.cn, b 551182097@qq.com, c 1462439214@qq.com, d 2964043664@qq.com *Corresponding author: lyanxing@nwnu.edu.cn Keywords: behavior detection, channel state information, k-means clustering algorithm Abstract: In order to improve the accuracy of indoor personnel detection, we propose a channel state information (CSI) indoor personnel behavior detection algorithm based on adaptive Kalman filter in this paper. After collecting the original data package of CSI, the adaptive Kalman filter algorithm of variance compensation is used to filter the original data, and the dichotomous K-means clustering algorithm is used to classify the collected data and establish the fingerprint database. Then the k-nearest neighbor (KNN) matching algorithm is used to match the real-time data with the fingerprint database data to achieve the indoor behavior detection. The experimental results show that compared with the LIFS and FIMD methods, the method can recognize the action behavior of indoor people more accurately. 1. Introduction With the precise demand of location service, indoor positioning system has become an increasingly hot technology field, and the indoor positioning method based on WiFi signal has drawn great attention to many researchers due to its openness and ease of use. As a kind of wireless network based on IEEE802.11 protocol, WiFi has been widely used in most families and office environments. Nowadays, most mobile devices have built-in wireless network card that conforms to IEEE802.11 standard, which makes it easy for users to access wireless local area network (WLAN), and its wide coverage significantly reduces the cost of indoor positioning technology. Indoor location technology has good development advantages in many fields, such as indoor intrusion detection, campus security, personnel detection in shopping malls, patient monitoring, real-time detection of the elderly and children at home [1]. Literature [2] proposes a low-cost and high-precision passive target location method lifs based on CSI model, which effectively applies the characteristics of CSI to target location, but this method does not consider the relationship between detection area and detection rate. In the reference [3], FIMD system uses the stability of CSI to achieve more fine-grained personnel detection in static environment, but it does not achieve high detection rate, and the system performance will be affected by the experimental environment. In the reference [4], the CSI signal is effectively reduced by sparse representation in frequency domain, Internet of Things (IoT) and Engineering Applications (2020) 5: 1-8 Clausius Scientific Press, Canada DOI: 10.23977/iotea.2020.050101 ISSN 2371-8617 1