Human Localization And Behavior Anomaly Detection Based On Multi-sensor Fusion

With the deepening of aging population degree and the changing of age distribution pattern, the traditional family pension mode faced great challenges, and the research of smart home care got the attention of scholars. Human localization and behavior anomaly detection, which are the basis of smart home care system, have been widely studied. However in the expect human localization the common algorithm is based on the pyroelectric infrared sensor network, and the accuracy of this algorithm is not high, and behavior anomaly detection has weaker anti-interference performance. In order to solve these problems, this article adopts the multi-sensor information fusion algorithm to detect human position and human behavior anomaly.In this thesis, the multi-sensor information fusion algorithm is proposed to improve the positioning accuracy, compared with the common algorithm, the algorithm in this thesis can obtain higher positioning accuracy. In the study of behavior anomaly detection, the character is found that different behavior status corresponding to different axial data of the acceleration sensor, based on the findings, the dual-threshold analysis method is proposed. It can eliminate the interference that running behavior acts on fall down detection effectively, and reach a higher anti-interference. In order to further improve the accuracy of behavior anomaly detection, lying down anomaly position detection algorithm is proposed, which can be treated as the supplement of fall down detection.The algorithms of human localization and behavior anomaly detection proposed in this thesis have been verified, and reached a higher detection accuracy. In this thesis, the research provides a platform for the intelligent pension system, but also laid foundation for the further study of smart home care, so it has highly practical value and wide research prospects.