Automatic Construction of a Minimum Size Motion Graph (2008 - 2009)

Motion capture data have been used effectively in many areas of human motion synthesis. Among those, motion graph-based approaches have shown great promise for novice users due to their simple graph structure, ability to generate long motions, and fully automatic synthesis of motions. The performance of a motion graph relies heavily on selecting a good set of motions to build the graph. This motion set needs to contain enough motions to achieve good connectivity and smooth transitions. At the same time, the motion set needs to be small for fast motion synthesis. Manually selecting a good motion set that achieves these requirements is difficult, especially given that motion capture databases are growing larger to provide a richer variety of human motions. Therefore we propose an automatic approach to select a good motion set. We cast the motion selection problem as a search for a minimum size sub-graph from a large motion graph representing the motion capture database and propose an efficient algorithm, called the Iterative Sub-graph Algorithm, to find a good approximation to the optimal solution. Our approach benefits novice users who desire simple and fully automatic motion synthesis tools, such as motion graph-based techniques.[Website]

Well-Connected Motion Graphs (2007 - 2008)

Motion graphs provide users with a simple yet powerful way to synthesize human motions. While motion graph-based synthesis has been widely successful, the quality of the generated motion depends largely on the connectivity of the graph and the quality of transitions in it. However, achieving both of these criteria simultaneously in motion graphs is difficult. Good connectivity requires transitions between less similar poses, while good motion quality results only when transitions happen between very similar poses. This paper introduces a new method for building motion graphs. The method first builds a set of interpolated motion clips, which contain many more similar poses than the original dataset. Using this set, the method then constructs a motion graph and reduces its size by minimizing the number of interpolated poses present in the graph. The outcome of the algorithm is a motion graph called a well-connected motion graph with very good connectivity and only smooth transitions. Our experimental results show that well-connected motion graphs outperformstandardmotion graphs across a number of measures, result in very good motion quality, allow for high responsiveness when used for interactive control, and even do not require post-processing of the synthesizedmotions. [Website]

Virtual Locomotion Controllers (2006 - 2007)

The ability to become fully-immersed in a 3D virtual world is currently limited by traditional human-machine interfaces such as mice, keyboards and game controllers. A more natural human-machine interface which uses sensorimotor responses that closely resemble the tasks and actions a user would physically perform in the real world would allow the user to better control their avatar in the virtual world, resulting in a fully-immersed experience. To address this need, an advanced human-machine interface, known as a Virtual Locomotion Controller (VLC), has been developed at soVoz Inc. The VLC combines solid state gyros and accelerometers, ultrasonic position sensors and force sensitive foot pads with real-time inverse kinematic techniques to provide these capabilities. As a result, the VLC provides an immersive user interface/game controller potentially usable with a wide range of application software (e.g. training, simulation, education, entertainment) and hardware platforms (PCs, embedded systems and game consoles). Possible applications include: commercial 3D computer games and virtual reality simulations, embedded dismounted infantry training systems, sports training and first responder/homeland security simulation systems.

Untethered Motion Capture Evaluation for Flightline Maintenance (2006 - 2007)

The purpose of this effort is to explore and evaluate the utility of novel motion capture technologies within the Air Force maintenance domain. A primary objective is to determine the potential of untethered motion capture capabilities for real-time human subject motion capture and performance data collection with full scale physical props. A resultant objective will be to evaluate data collected during maintenance task performance validation for the purpose of instruction generation, and maintenance training. This effort will consist of a domain analysis, a conceptual design definition, a prototype development, and a performance evaluation within relevant operational maintenance scenarios. Program objectives will be achieved by using both university laboratory and field based research to evaluate the efficacy of untethered motion capture for obtaining human performance data from tasks involving full scale physical props, and the reuse of this information within augmented procedural instructions.

American Sign Language Generation and Evaluation (2005 - 2007)

The goal of this project is to develop new technologies that enable the machine translation of English text into animations of American Sign Language (ASL). This research will make more information and services available to the majority of Deaf Americans who face English literacy challenges. Because signed languages, like ASL, contain phenomena not seen in traditional written/spoken languages, they are particularly challenging to process using standard machine translation (MT) approaches. Exploring the computational linguistics of ASL can help us understand the limitations of current MT technologies and motivate the development of new ones. [Website]

LMCO: Virtual Human Testbed (2004 - 2005)

Simulating human reach is still challenging when considering complex interactions with the environment. Standard approaches involve inverse kinematics (IK) methods and usually require a complete but exponential cost search in configuration space. In ergonomic applications, both "naturalness" and interactive performance are important. We describe a real-time, collision-free, sternum-rooted IK solution for an articulated human figure based on motion capture data, human strength models, and multi-joint coordination functions. Movement paths are discovered through spatial search in a partitioned workspace. The system generates natural collision-free reach motions in realtime. The resulting animations and statistics demonstrate the efficacy of this approach. [Video, 22.3MB]

Some Interesting Class Projects I did.