Biography
Enrollment Date: 2013
Graduation Date：2016
Degree：M.S.
Defense Date：2016.05.31
Advisors：Chun Zhang Sheng Zhang
Department：Institute of Microelectronics，Tsinghua University
Title of Dissertation/Thesis：Design of Motion Capture Device Based on INS Techniques
Abstract:
The past decade has witnessed the rapid development of a series of emerging and practical artificial intelligence techniques represented by virtual reality (VR), which raises more new requirements for human-computer interaction (HCI) techniques. As a result, efficient, accurate and convenient HCI techniques are those that conform to the trend. Motion capture (MoCap) is widely used in military and athletic sports training, rehabilitation, recreation and other scenarios as the HCI. Existing MoCap techniques are mostly vision-based, depending on image or video. Optical measuring systems of high-precision are always expensive and too sophisticated, while the inexpensive products can hardly satisfy the increasing need for MoCap performance. Compared with optical devices, inertial sensors based on microelectromechanical-systems (MEMS) technique are more cost-effective, and much easier to operate, which makes them less sensitive to environmental influences and thus more suitable for product use. This paper proposes a design of MoCap device based on inertial-navigation-system (INS) techniques, and the soft- and hardware schemes are determined based on research. The MoCap system proposed in this paper uses field-programmable-gate-array (FPGA) as the processor hardware of the embedded subsystem. One system-on-a-programmable-chip (SoPC) built in FPGA is used to organize the processing unit and various peripherals. A re-designed Avalon-compatible data interface module which transfers data in a creative way is embedded in the SoPC. It improves the efficiency of data-transmission and takes advantage of the higher multi-channel data throughput of FPGA. The solving program of attitudes are based on an optimized gradient descent algorithm, which is implemented in both the embedded and the host subsystem, either of which can be selected to use in different applications. The hardware driver and data pre-processing approaches are programmed and built into classes according to the object-oriented programming theory. Finally the features are packaged into application-programming-interface (API) in the form of dynamic-link library (DLL), to be used in other programs. The most sophisticated part of a MoCap system, data glove, is chosen as the experiment carrier, and a complete platform for verification is developed. Besides, the method for collecting data which forms in this design is creatively applied to the tactile detection in teleoperation control of manipulators. For the sake of making usable device, the structure model of humans' hands are researched and the constraints and mapping among sensor nodes are determined, followed by the development of 3D render model. The 3D demo application is developed in the environment of Microsoft Windows, using a 3D graphic engine called OGRE. Finally we accomplished the experiments for validation. Results show that the ideas and approaches proposed in this paper are reasonable and also practical.