Human Motion Capture And Recognition Based On Body-Sensing Network

With the increasing level of technology and social demand,motion capture and recognition technology have been widely used in many fields such as the healthcare industry,film industry,and sports industry.Compared with visual motion capture,body-sensing-network-based human motion capture and recognition has the advantages of low latency,high flexibility,and good portability.As a result,its research significance and commercial value have attracted attention both from academic and commercial communities.Existing body-sensing network systems are generally expensive,difficult to use,and close to extensions.This paper aims to design a motion recognition and visualization platform that is cost-effective,energy-efficient,and easy to use,and builds a body-sensing-network-based motion capture and recognition system with Wi-Fi communication technology.First,this paper designs and implements a wireless low-power body-sensing network system based on Wi-Fi technology.After analyzing the requirements of the system following the top-down principle,overall scheme design and module division of the body-sensing network system are completed,and the software and hardware subsystems are designed respectively.The hardware subsystem includes circuit schematic design of data collection nodes and wireless sensor nodes,PCB designing,and transplantation of Linux for embedded systems.The software subsystem is composed of the communication module,the data collection and storage module,and the wireless sensor node management module.Then,this paper proposes a position and attitude calculation algorithm for a single node.The nine-axis attitude is calculated by fusing data from multiple sensors using the complementary filtering algorithm.The experimental results show that the accuracy and speed of the proposed method are like those from commercial sensors.To reduce the influence of the integral error on the accuracy of the position calculation,this paper reconstructs the position and velocity based on the object space tracking model.The experiment verifies that this method improves the accuracy of the position calculation.After that,this paper introduces technologies used to store,manage,and visualize human body movements.Raw data of human body movements from the body-sensing network is transformed to the Biovision Hierarchy format for storage and is parsed and visualized using Java Script and a library called three.js.Experiments indicate that the method is effective and flexible in the storage and visualization of human body movements.To improve the usability of the system,this paper also proposes a management system to manage data such as body-sensor network nodes,motion data,and human skeleton models.Finally,this paper proposes a DTW-k NN based method to fuse data from multiple sensors to identify gait phases.In this paper,the ISDTW algorithm is used to avoid the singularity problem of the DTW algorithm.Compared with previous methods,the ISDTW algorithm does not have parameters and improves the effect of the DTW similarity measurement.According to the characteristics of the k NN clustering algorithm,pruning is applied to reduce the number of DTW operations in gait phase recognition and the time complexity of the overall algorithm.Experiments show that the ISDTW-k NN algorithm proposed in this paper has significant improvements in accuracy and real-time performance compared with DTW-k NN,and it also proves that the DTW and k NN algorithms can effectively recognize human body actions.This paper designs and implements a body motion capture and recognition system based on the body-sensing network through the following steps: hardware system production,wireless communication protocol design,pose calculation,motion visualization,and gait phase recognition.The experimental results show that the system designed in this paper achieves the goals of cost-effective,energy-efficient,and easy to use.The system can be potentially used in the fields of standardized sports training and patient rehabilitation.