8/14/2019 Ph.D. Candidate in Computer Science, Expected Graduation

1/3

Yi Sun Ph.D. candidate in Computer Science, expected graduation date: Summer 20106439 166th PL SE, Bellevue WA 98006, E-mail: ysun@cs.binghamton.edu, Phone: (503)481-8374

Objective A technical position in building intelligent systems using machine learning, pattern recognition,computer vision, or image processing techniques.

Qualifications Machine learning, pattern recognition, data mining, computer vision and image processing;Languages: C/C++/Java/MATLAB, Scheme, Prolog, and Haskell;Operating systems: windows and unix;Knowledgeable in OpenGL and OpenCV;Excellent written and oral communication skills.

Experience Research Assistant Graphics and Image Computing Lab, SUNY atBinghamton

Binghamton, NY 13902 2005-present1. Designed and developed a real-time face detection and eye pupil detection system using livestream video cameras/web cameras. This system captures the illumination invariant features offacial and eye pupils and computes mutual information of local candidate eye patches to find the

best eye pupil candidate location in almost real time (more than 10 fps) and the detected eyepupil location is more than 95% correct within 3 pixels error tolerance. The system is developedunder Eclipse C++environment and uses OpenCV.2. Designed and developed a 3D face recognition and 3D facial expression recognition system.This system extracts geometric/temporal features from a high volume (1.5 TB) facial data baseand applies several machine learning algorithms (2D HMM, genetic algorithm, Bayesian, anddiscriminative analysis like PCA/LDA) for recognition tasks. We demonstrated our systemhas better recognition performance compared to traditional 2D image/video based approaches.(Using C++, MATLAB)3. Designed and developed a wrl/obj reader software that can be a post-processing tool to load,visualize, render, and edit 3D dynamic and static facial models. This tool can read structuredfiles like wrl or obj files to construct 3D shape model and supports standard bmp or jpg file as

texture file for 3D model rendering. Its developed under Microsoft Visual Studio 2003 and usesOpenGL.4. Designed and developed an automatic 3D head pose orientation estimation algorithm. Thisalgorithm uses unsupervised learning method to find the facial feature that is robust with poseand expression variations and achieves 97% correct estimation. Its developed using C++andMATLAB.

Research Assistant ECE Department, Oregon Graduate InstituteBeaverton, OR 97006 2002-2003Worked on computer vision and medical image processing. Designed and developed an activeshape based image segmentation algorithm for prostate contouring of QCT images. This algo-rithm first removes the noise in QCT images by applying low band filters then uses active shapecontouring method to find the candidate prostate contour in the edge image of the original QCT

image. The accuracy is 95% compared with manually labeled contour by experts. This algorithmsaves doctors time from five minutes to less than two seconds for contouring possible prostatetumor shape in one QCT image. Its developed under Microsoft Visual C++6.0 environment.

Software Development Engineer UAES, a joint venture of Robert Bosch GmbH,Germany

Shanghai, China 201206 2002Developed software on Electronic Control Unit (ECU) modules for automobiles, including im-plementation, debugging and performance test. Developed a real-time temperature monitoringsoftware that displays and communicates with remote single chip devices using C and C++)

Research Assistant Inst. of Image Processing & Pattern

Recognition, Shanghai Jiaotong UniversityShanghai, China 200030 1999-20021. Designed and developed a heart-beat fault detection module and an interactive visual com-munication module for a remote patient monitoring system. The system is composed of bothclient and server ends. The client end is connected with sensors and has a web camera to mon-itor the patient. The server end receives the ECG signal and monitoring video from the client

8/14/2019 Ph.D. Candidate in Computer Science, Expected Graduation

2/3

simultaneously, detects wave pattern from electrocardiography (ECG) and reports alarm whenthe detected wave pattern has significant difference with store normal patterns. Once the alarmis on, the operator in the server end can control the cameras setting remotely to have a betterview of the patients situation.2. Developed a sequential image registration algorithm based on the spatial distribution of ob-jects of interest, this algorithm helps find the best match location of the query image from thecurrent image by 20 percentage.3. Integrated an image processing tool with functionalities including image enhancement, colorspace transformation, geometric transformation, histogram equalization, and image registration;

Education State University of New York (SUNY) at Binghamton, NY 01/2004 - presentCandidate of Philosophy of Doctor in Computer ScienceAdvisor: Professor Lijun YinPh.D. thesis topic: Spatial/temporal 3D facial surface analysis for recognition.Traditionally, human facial expressions have been studied using either 2D static images or 2Dvideo sequences. The 2D-based analysis is difficult to handle large pose variations and subtlefacial behavior. This exploratory research targets the facial expression analysis and recognitionin a 3D space. The analysis of 3D facial expressions will facilitate the examination of the finestructural changes inherent in the spontaneous expressions. My thesis work aims to achieve ahigh rate of accuracy in identifying a wide range of facial expressions, with the ultimate goal of

increasing the general understanding of facial behavior and 3D structure of facial expressions on adetailed level. I developed a curvature distribution based algorithm on static 3D static databaseand conducted comparative study with 2D image based approaches for facial expression recogni-tion. I also developed a real 2D HMM plus curvature transformed LDA based algorithm for facialexpression recognition and designed different structures of classifiers to fit the dynamic 3D facescenario and other related projects. The study demonstrated that the combination of spacialand temporal dynamics of humans 3D facial surface achieves better recognition performancecompared with the 2D and static 3D approaches.

Oregon Graduate Institute, OHSU, OR 08/2002 - 12/2003Master of Science in Electrical EngineeringArea: machine learning based image segmentation

Shanghai Jiaotong Unversity, Shanghai, China 09/1999 - 03/2002Master of Science in Pattern Recognition and Intelligent SystemsThesis: Sequential image registration based on selective attentionShanghai Jiaotong Unversity, Shanghai, China 09/1995 - 06/1999Bachelor of Science in Information Detection and Control (major)Bachelor of Science in Computer Science and Engineering (minor)

Awards andScholarships

Graduate Student Awards of Excellence in Research, SUNY at Binghamton, 2009.Travel grants from the European Conference on Computer Vision (ECCV), 2008.Graduate with honor of Shanghai (Top 1%), China, 2002.Guanghua Scholarship (Top 2%), Shanghai Jiaotong University, Shanghai, China, 2000 and 2002.

Siemens Scholarship (Top 1%), Shanghai Jiaotong University, Shanghai, China, 2001.Zhu Chuan Qu Award (10 awarded annually), Shanghai Jiaotong University, 2001.Ren Min Scholarship, Shanghai Jiaotong University, 1995-1996.

SelectedPublications

1.Y. Sun and L. Yin, Evaluation of spatio-temporal regional features for 3D face analysis, IEEECVPR09 Workshop on Biometrics, Miami, Florida USA, June 2009.2. Y. Sun and L. Yin, Automatic pose estimation of 3D facial models, 19th conference of Inter-national Conference on Pattern Recognition (ICPR08), Tampa, Florida, USA, December 2008.3. Y. Sun and L. Yin, Facial expression recognition based on 3D dynamic range model sequences,10th European Conference on Computer Vision (ECCV08), October, 2008, Marseille, France.4. Y. Sun, M. Reale and L. Yin, Recognizing partial facial action units based on 3D dynamicrange data for facial expression recognition, 8th IEEE conference on Automatic Face and GestureRecognition (FG08), September, 2008, Amsterdam, the Netherlands.5. Y. Sun and L. Yin, 3D spatio-temporal face recognition using dynamic range model sequences,IEEE CVPR08 Workshop on Biometrics, Anchorage, Alaska, USA, June 2008.6. Y. Sun and L. Yin, Genetic algorithm based approach for 3D face recognition using geometricface modeling and labeling, (Part I: Biometrics, Chapter 4), p95-118, 3D Imaging for Safety andSecurity, A. Koschan, M. Pollefeys, and M. Abidi (editors), Computational Imaging and Vision

8/14/2019 Ph.D. Candidate in Computer Science, Expected Graduation

3/3

Series, Vol. 35, Spinger, 2007. (Book chapter)7. Y. Sun and L. Yin, Evaluation of 3D Facial Feature Selection For Individual Facial ModelIdentification, 18th conference of International Conference on Pattern Recognition (ICPR06),Hong Kong, Aug, 2006.8. Y. Sun and L. Yin, 3D face recognition using two views face modeling and labeling, IEEECVPR05 Workshop on Advanced 3D Imaging for Safety and Security (A3DISS), San Diego, CA,2005.9. Y. Sun and L. Yin, A genetic algorithm based feature selection approach for 3D face recogni-tion, Biometrics Symposium, Washington DC, 2005 (sponsored by NIST, DOD, NSF, etc.)

Activities IEEE memberConference review: International Symposium on Visual Computing (2006 and 2007); FirstIEEE Workshop on Face Processing in Video (in conjunction with IEEE CVPR04); ICPR (2006and 2008); FG08; IEEE Third International Conference on Biometrics: Theory, Applicationsand Systems (2009).Journal review: Pattern Recognition Journal (2007); International Journal of Pattern Recog-nition and Artificial Intelligence (2007); IEEE Transactions on Pattern Analysis and MachineIntelligence (2009).