Individual Soldier Activity Recognition And Positioning System Based On Inertial Sensors

With the continuous development of multi-sensor technology,information fusion tech-nology,computer technology and other scientific technologies,the modern war mode has grad-ually evolved from traditional war to today's information war.The individual soldier combat system is an important manifestation of modern warfare systems,and it is very important for information-based warfare.The human behavior recognition and positioning are important parts of the individual soldier combat system.The individual soldier behavior recognition and positioning system studied in this sub-ject is built with low power consumption and portable hardware.It introduces deep learning methods to recognize human behavior and obtains good results in actual tests.The Beidou navigation system with inertia navigation can enables 24/7 individual positioning.This article first designs individual soldier nodes and server nodes.Individual soldier nodes mainly include inertial sensors,local information processing platforms and display mod-ules.They are mainly designed for completing the collection and data processing of individual soldier information,as well as local real-time behavior recognition and positioning? The server node is mainly designed for monitoring the status of each individual soldier node in real time.The software part designs the wireless communication method between the individual soldier node and the server node,and designs and develops the individual soldier monitoring interface,which can realize the real-time graphical monitoring of the status of each individual soldier.Then,the research of human behavior recognition based on inertial sensors is carried out.In terms of data preprocessing,the offset of the original data is first corrected,and the median filtering method is used to suppress the noise.The acceleration of gravity is filtered to obtain linear acceleration which is more useful.The attention-based LSTM(Long Short-Term Mem-ory)network for human activity recognition is proposed,which is verified by experiments to obtain very excellent accuracy on the Opportunity dataset.Finally,the algorithm deployment of deep learning on the Raspberry Pi is studied,and actual experiments are performed.The experimental results show that the human behavior recognition system has a high accuracy rate,which can meet the needs of the subject.In the end,this article conducts the research on individual soldier positioning.This topic uses dead reckoning and zero speed correction algorithms to perform individual soldier po-sitioning without Beidou navigation signals,and describes the attitude calculation algorithm,speed position update algorithm and zero speed correction algorithm used? Several linear and circumferential positioning experiments have been performed based on inertial measurement unit LSM9DS0.Using Beidou navigation system with inertial navigation can achieve 24/7individual soldier positioning.