• Virtual Humans
• Machine Learning
• Computational Emotions
• Intelligent Tutoring Systems
• Cognitive Science

Virtual Humans The Virtual Human Project aims to develop human-like software agents for virtual training environments. To create believable virtual humans, we are conducting research on areas such as computational emotion modeling, nonverbal behavior generation, and natural language processing and generation. Embeded Stanford Parser into Virtual Human system-SASO A natural language parser is a program that works out the grammatical structure of sentences, for instance, which groups of words go together (as "phrases") and which words are the subject or object of a verb. Probabilistic parsers use knowledge of language gained from hand-parsed sentences to try to produce the most likely analysis of new sentences. These statistical parsers still make some mistakes, but commonly work rather well. Their development was one of the biggest breakthroughs in natural language processing in the 1990s. The Stanford Parser package is a Java implementation of probabilistic natural language parsers, both highly optimized PCFG and lexicalized dependency parsers, and a lexicalized PCFG parser. Annotated Multimodel Corpus of Nonverbal behaviors Despite the spread of software development and software usage on computational facial expressed emotion, we have only a little data of computational gesture emotion.  In real time, Non-verbal behavior, which is the gesture movement, as well as the facial expression, plays an important role in detect people’s emotion. Our main proposal is to set 3D gesture models for Emotion and use it in real time multi-agent system.    Then we will compare the real time date with the 3D emotion model training set to justify the emotion type.  The usage of this algorithm might be on 3D visual game, military negotiate training, education, safety check on airport, and etc. Published Paper   Jina Lee,  Zhiyang Wang,  Stacy Marsella.  Evaluating Models of Speaker Head Nods for Virtual Agents. The 9th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2010), Accepted.

USC Institute for Creative Technologies USC Cognition and emotion Group The Stanford Natural Language Parser AAMAS 2010, which is the top conference on AI-Agent field Supervised by Professor Stacy Marsella Colleague Jina Lee

Course Description:     This course introduces students to physically based simulation for computer graphics and related fields. Physically based simulation is an active research area in computer graphics, with applications to computer games, virtual reality systems, and movie special effects. Efficient numerical methods for simulating a variety of visually interesting physical phenomena will be discussed in the context of interactive simulation. Topics include deformable objects (solids, cloth), fluids, sound simulation, collision detection, haptics, rigid body dynamics, and GPU programming. In addition to computer science students, this course should also be appropriate for graduate students in related disciplines such as mathematics and physics. Project Description: Simulating a Jello Cube  This project is to do a physically-based simulation of a jello cube. The jello cube is made of elastic material.  When the jello cube is stretched, it will try to contract. When squeezed together, it will tend to inflate back to the original shape. Inorder to simulate the jello, I used a 3D mass-spring network to achieve the goal.  Coded with C++ language and OpenGL library

USC Computer Science Department USC CSCI599 Physically Based Modeling for Interactive Simulation and Games Supervised by Professor  Jernej Barbic

Project Description: The course covered the physics of MR, selective excitation, image acquisition, image contrast, volumetric imaging, and various system imperfections; and then covers image advanced topics related to entrepreneurship, rapid imaging, RF pulse design, and image artifact correction.    Coursework was motivated by clinical and research applications such as cardiac imaging, flow measurement, and functional MRI.   Focused on   Reconstructed Partial and Non-Cartesian k-space reconstruction for MRI Image  Coded with Matlab language

USC Electrical Engineering(Multimedia and Creative Tech) USC EE591 Megnetic Resonance Imaging and Reconstruction Supervised by Professor  Krishna Nayak

 

It is a National Institutes of Health-supported interdisciplinary project, involving researchers from the fields of Communication, Cell Neurobiology, Computer Science, and Physical Therapy. The purpose of the project is to develop virtual therapeutic environments that include different levels of haptic sensory feedback for post-stroke rehabilitation. Various applications have been created within virtual environments using the PHANToM and CyberGrasp (haptic devices), ranging from everyday functional tasks to game-like activities designed to motivate patients and to maximize therapeutic movement with cortical reorganization goals in mind.  Methodologies are developed to diagnose patient's current status and evaluate his/her progress of the function of upper extremity via clinical test data.       Focused on Developing an infer array sensor interactive system for optical tracking devices through which users interact with 3D games.  Coded with C++ language (OpenCV , OpenGL) and model based algorithm.       View DEMO

USC Institute for Creative Technologies USC Medicine Department ISNSR Group Supervised by Luke Yeh, Lead by Professor Alexander A. Sawchuk Colleague Song-Hua Xing

 

Project Description: Students are expected to comprehend existing C/C++ programs and modify the codes for various goals. The students may also be required to write small programs from scratch.  Provided sample codes come with makefiles for compilation under Unix environment.  Either familiarity with basic Unix commends or the ability to covert the codes to a windows project is required. Focused on   Implement Huffman, LZW, Run-length method of data compressing    Implement QM coder, Lloyd-Max Scaler Quantizer and Vector Quantizer Design            Simulating various modes of encoding-decoding with JPEG    Coded with C++ and Matlab language and compiled on Unix.

USC Electrical Engineering(Multimedia and Creative Tech) USC EE669 Multimedia Data Compressing Supervised by Professor C.-C. Jay Kuo

Project Description: This is supplemental course information, designed to give you a fuller picture of the course and an expanded look at the topics covered.  Image sampling, 2-D image transform, image enhancement, geometric image modification, morphologic processing, edge detection, texture analysis, image filtering and restoration Focused on   Implemented the fix threshold Digital halftoning, dithering matrix and Floyd-Steinberg's error diffusion with serpentine scanning to convert lena.raw to a binary image                                   Applied edge detection algorithms with multi-order derivative method           Coded with C++ and Matlab language

USC Electrical Engineering(Multimedia and Creative Tech) USC EE569 Digital Image Processing Supervised by Professor C.-C. Jay Kuo

Project Description: Grid format is a common home and abroad DEM data formats, DEM Grid data format also known as digital elevation model data grid, that space will be separated into rules-based grid, the grid in the corresponding property value is given to a geographical entity that data forms of organization. The objective is to issue Grid format data elevation model data to coordinate data and elevation data extraction and data extracted from the visual display reasonable. Focused on   Designed the  whole system  Analyzed DEM image data with ArcGIS, transforming the DEM image data into a binary image  Coded with C++  language and ArcGIS Desktop

Beihang University (the Beijing University of Aeronautics and Astronautics) Supervised by Huaping Xu

Project Description: Teradata is a division of NCR Corporation. We put the unbridled power of the world's leading enterprise data warehouse solution into our customers' hands. Our solutions deliver a great ROI through a simple and efficient infrastructure, built for speed and infinite growth. And that frees our customers to uncover new opportunities for business growth, increased efficiency, and improved customer relationships. Focused on   Acted as a Software Quality Assurance Intern       Manually tested database application software and contributed to automation testing   Managed daily testing logs for whole testing group

NCR-Teradata Company Supervised by Jun Li

For more information, please see the resume page.