THESIS DEFENSE Department of Computer Science and Engineering, University of South Carolina Candidate: Surya Bhatt Advisor: Dr. Csilla Farkas Date: Tuesday, March 24, 2015 Time: 2:00pm Place: Swearingen (3A75) Abstract There’s a significant increase in online consumer forums. When customers set out to buy a product they have a preconceived idea about it. They mostly form the idea/expectation from print, audio/video advertisements, online reviews and word of mouth. There have been previous researches on how user generated content can be used to determine the market structure insights [Meeyoung Cha et. al. 2007]. The audio analysis is divided into two parts. One is linguistic research which focuses on the language and content of the input. The second one is paralinguistic which focuses on how words are spoken (prosodic effects). Our research focuses on comparing Brand positioning maps based on consumer reviews with and without linguistic audio analysis of advertisements and reviews. Our goal is to show that this hybrid approach helps us analyze the effectiveness of advertising on brand positioning maps. This approach shall also help us in making association graphs for a brand using words of perception/opinion flowing with that brand/product. Which may in turn assist companies in improving the focus of their advertisements in order to persuade the required set of crowd and drive the public perception.

DISSERTATION DEFENSE Department of Computer Science and Engineering, University of South Carolina Candidate: Ping Liu Advisor: Dr. Yan Tong Date: Monday, March 23, 2015 Time: 2:30pm Place: Swearingen, 3A75 Abstract As being characterized by various configurations of facial muscular movements, facial expression is the most natural and powerful means for human communications. A robust and accurate facial expression recognition system, which is with capabilities to automatically recognize facial activities in given images/videos, has applications in a wide range of areas. However, developing such an automatic system encounters several challenges. As a standard pattern recognition problem, facial expressions recognition (FER) consists of three major modules in training: feature learning/extraction, feature selection, and classifier construction. This research aims to improve the first two modules respectively, and furthermore integrate them in an iterative and unified way to enhance the final recognition performance. In order to improve the performance of feature learning, novel log transformed sparse coding features with a spatial pyramid structure are proposed to characterize nonrigid facial muscular movements with head movements. To select the most distinctive features in recognizing facial expressions, a framework based on kernel theory is proposed to choose facial regions and analyze their contributions to different target expressions. Moreover, to unify the three training stages, a model taking advantage of both deep learning and boosting theory is proposed, with the capability of learning hierarchical underlying patterns that describe the given images and automatically selecting the most important facial regions for facial expression analysis. The proposed methods have been validated on public databases including spontaneous facial expression database, which is collected under realistic environment including face pose variations and occlusions. The experimental results show that the proposed methods yield significant improvements in recognizing facial expressions. More than that, the significant performance improvement in cross-database validation demonstrates the good generality of our proposed methods.

DISSERTATION DEFENSE Department of Computer Science and Engineering, University of South Carolina Emerging Wireless Sensor Networks in Intertidal Zones and for Tracking Targets Candidate: Miao Xu Advisor: Dr. Wenyuan Xu Date: Monday, March 23, 2015 Time: 10:00am Place: Swearingen (3A00, Deans Conference Room) Abstract Wireless sensor networks (WSNs), which consist of a large number of low-power, low-cost sensor nodes, are promising because they can continuously and remotely monitor our surroundings with few human involvements. For example, we have witnessed sensor applications that monitor the habitat environment of wild animals or monitor the health condition of infrastructures. These traditional WSNs are typically used as monitoring applications and deployed at various terrestrials. In this dissertation, however, we examine two different types of 0emerging sensor applications, which are WSNs deployed in intertidal zones (IT-WSNs) and WSNs deployed for tracking targets (TT-WSNs). Alternately submerged and laid bare by tides, the wireless channels in IT-WSNs are intermittent and could be unavailable up to 24 hours, which makes IT-WSNs unique compared with terrestrial WSNs. In the first part of this dissertation, we examine some problems that are caused by the influence of tides. First, we study the time synchronization problem in IT-WSNs. Due to the intermittent wireless channels, traditional resynchronization schemes cannot be directly applied to IT-WSNs. To prolong the resynchronization interval, we propose a temperature-aware compensation scheme that continuously adjusts the clock locally without any resynchronization messages. Second, we examine the low-power embedded systems for the sensor nodes in IT-WSNs. Battery powered, the nodes in WSNs typically employ duty-cycling schemes to reduce the average power. To further reduce the power consumption, we study the channel-aware features for radio transceivers given that the channels are intermittent. We also propose to develop an automation platform that automatically finds the most power saving configuration of IO pins. Finally, we investigate environment-aware MAC protocols in Part I. Many existing WSNs are proposed for monitoring physical phenomena, but few of them study the applications for tracking targets. In the second part of this dissertation, we examine the TT-WSNs that are deployed for tracking a target such as some endangered animals. Since the data exchanged in TT-WSNs are typically associated with the target's location information, the data privacy is important to the users. In addition, once deployed, the networks are typically left unattended, which makes the privacy issue even more challenging. Although many cryptography-based solutions have been proposed to solve the problem, we are interested in non-cryptography schemes because cryptography-based methods cannot cope with node compromise. Based on the observation that message content is important in TT-WSNs, we propose a content-aware data dissemination scheme for TT-WSNs to achieve better data privacy and higher data availability with less energy cost.

WIC meeting. March 5th, 6pm, room 3A75: Carol Wallace, Kate Ingram, and Tracy Cobb from Blue Cross Blue Shield. More information

This Friday, Ioannis Rekleitis from the Department of Computer Science and Engineering will be speaking at 2:30 - 3:20 pm in LeConte 312. Abstract The last few years, robots have moved from the pages of science fiction books into our everyday reality. Currently, robots are used in scientific exploration, manufacturing, entertainment, and household maintenance. While the above advances were made possible by recent improvements in sensors, actuators, and computing elements, the research of today is focused on the computational aspects of robotics. This talk presents an overview of algorithmic problems related to robotics, with the particular focus on increasing the autonomy of robotic systems in challenging environments. In particular I would discuss the use of discrete structures such as graphs to efficiently solve robotic problems. Cooperative Localization Mapping and Exploration employs teams of robots in order to construct accurate representations of the environment and of the robots' pose. The problem of coverage has found applications ranging from vacuum cleaning to humanitarian mine removal. A family of algorithms will be presented that solve the coverage problem efficiently in terms of distance travelled. Finally, I will present some current work on the problem of searching under uncertainty. The work that I will present has a strong algorithmic flavour, while it is validated in real hardware. Bio Ioannis Rekleitis is currently an Assistant Professor at the Computer Science and Engineering Department at the College of Engineering and Computing, University of South Carolina. Previously he was an Adjunct Professor at the School of Computer Science, McGill University. Between 2004 and 2007 he was a visiting fellow at the Canadian Space Agency. During 2004 he was at McGill University as a Research Associate in the Centre for Intelligent Machines with Professor Gregory Dudek in the Mobile Robotics Lab (MRL). Between 2002 and 2003, he was a Postdoctoral Fellow at the Carnegie Mellon University in the Sensor Based Planning Lab with Professor Howie Choset. He was granted his Ph.D. from the School of Computer Science, McGill University, Montreal, Quebec, Canada in 2002 under the supervision of Professors Gregory Dudek and Evangelos Milios. Thesis title: "Cooperative Localization and Multi-Robot Exploration". His Research has focused on mobile robotics and in particular in the area of cooperating intelligent agents with application to multi-robot cooperative localization, mapping, exploration and coverage. His interests extend to computer vision and sensor networks. He has worked with underwater, terrestrial, aerial, and space robots. Ioannis Rekleitis has published more than sixty journal and conference papers. His work can be found online at: http://www.cse.sc.edu/~yiannisr/

More information here.

THESIS DEFENSE Department of Computer Science and Engineering, University of South Carolina Planning A Virtual Lab for Analysis of Malware Candidate: Subhro Sankha Kar Advisor: Dr. Marco Valtorta Date: Friday, February 20, 2015 Time: 10:00 am Place: Swearingen (3A00, Dean’s Conference Room) Abstract I will present a study of the development and availability of different virtual infrastructure platforms and methods of virtualization for the Intel architecture. I will discuss various approaches to deployment and management of a virtual lab that can be used for the study of operating systems and the analysis of malware. My approach is to deploy a para-virtualized analysis lab that is functionally equivalent to the Red Hat malware analysis lab and that uses open source software. I will show how I completed this task using OpenStack, a platform that was not designed for malware analysis. I will present instructions for the deployment and management of such a virtual infrastructure, compare its cost to that of a full hardware lab, and show how the para-virtualized lab overcomes countermeasures taken by a typical piece of malware when running in a virtualized environment