Research Of Real-time Human Motion Capture System Based On Inertial Sensors

Motion capture is a multidisciplinary and an application field combined with the interaction of technology developed in recent years, involving electronics, communication, control, computer graphics and human ergonomics, navigation etc. different subjects. It is currently a hot spot in the field of human-computer interaction and has a broad development prospects in the film and television production, the robot control, interactive games, sports training, medical rehabilitation, etc. and other fields. In the field of game design, many large sports games have combined human motion capture with virtual reality technology, making the objects more vivid and lifelike, such as the "3Dbasketball", "NBA". In film industry, many films, such as "Avatar", "The Adventures of Tintin", use human motion capture and virtual reality technology widely, making the film screen more infectious; while in the field of medical science, the technology is widely used in department of orthopedics, surgery operation, physical therapy and other fields, which can help doctors to better understand patient’s condition through comparing the motion attitude of patients before and after the rehabilitation. In sports training field, real-time monitring of athletes training state is realized through the display of real-time tracking interface by athletes wearing motion capture equipment.In this article, so as to select the appropriate sensors, we conducted the research by studying the finger motion in 3D space, and compared the working principle and performance parameters of the sensors, furthermore, it carried on the analysis of the possible error of each sensor. For example, it made a detailed analysis of the data jitter and skew of the acceleration sensor, the accumulated error of gyroscope, and putted forward the effective solution. It also made detailed schematic design for the sensor node and sink node which deployed in the body nodes, and designed the system data transmission protocol. At the same time, it carried on the analysis to the deployment location of sensor nodes in body nodes, and completed the fusion processing of multi sensor data based on the basis of inertial navigation theory, finally the attitude information was outputted by quaternion. In addition, through the analysis of hand biological structure, which built the virtual human motion model based on three-dimensional perspective by combining with the VS2012 development environment in the OpenGL graphics library and 3DS MAX software, at last, we realized the capture and reproduce of the movement gesture through the virtual model initialization calibration and data driven by the attitude data calculation.It proved that the human motion capture system has stable and reliable working performance through the experiment, which can provide new references for future research and design of motion capture system, also it has certain innovation and application value.