The Smart Clothing System Based On Digital Twin And Sensor Information Calculation

With the development of social information technology,various smart wearable devices have emerged.The existing wearable products such as smart bracelets,watches,and glasses on the domestic and international markets are small in size,but the location of the sensor arrangement is therefore limited,and the integration of sensors into clothing can solve this problem well.An important function of wearable devices is to identify the movement status of the wearer,and the existing sensor-based human behavior recognition technology cannot meet the needs of both high accuracy and real-time detection.Therefore,researching and designing a human action recognition algorithm with high accuracy and real-time and applicable to various edge computing platforms can solve this technical pain point.As an emerging research area of smart wearable devices,smart clothing has unprecedented opportunities.In this research context,this paper designs and builds a smart clothing system using multimode sensors,artificial intelligence technology,and digital twin technology.At the same time,the human action recognition algorithm in this paper is researched and improved,which makes the detection of user actions accurate and real-time,and builds a real-time monitoring platform.The main work and results of the paper are as follows.1.The use of multi-mode sensors integrated into the clothing,built a Raspberry Pi-based smart clothing hardware platform.The sensors used in this study are: MAX30102 heart rate sensor,MAX90614 infrared body temperature sensor,ATK1218-BD Bei Dou GPS dual-mode positioning module,and WTGAHRS2 10-axis inertial navigation attitude sensor.These sensors are used to collect user behavior information,including heart rate,blood oxygen,body temperature,positioning information,three-axis acceleration,and angular velocity information.This multimodal information can be used to accurately determine the user’s state of motion.2.The human action recognition algorithm based on multimodal sensor information is studied,and two human action recognition networks based on the fusion of high-dimensional information and spatial-temporal features,TS-HAR and ST-HAR,are proposed.These two networks combine the feature vectors of sensor signals in the time and frequency domains to extract the global temporal and spatial information of sensor signals.This paper mainly improves the spatial-temporal feature extractors:(1)the ATB of the transition layer based on the attention mechanism is proposed in TS-HAR,which improves the network recognition accuracy to 96.34%;(2)the TFE of the temporal feature extractor based on the continuous coding module is proposed in ST-HAR,which enables ST-HAR to obtain the global temporal features of the sensor signals and the accuracy is thus further improved to 97.01%.3.Based on the digital twin technology,a 3D digital twin monitoring platform that can map the user’s action state in real time is built.This study uses digital twin technology to display the user’s state in a three-dimensional form in the PC interface by increasing the interaction between sensors and digital twins,and realizes the real-time mutual mapping of the user’s state from physical space to digital space.