The human eye detecting and tracting in Driver fatigue monitoring

Tang Wei 공주대학교 대학원 2009 국내석사

(An) Adaptive QoE Management Scheme on Hybrid Remote Display Protocol for Mobile Thin Client Computing

Wei Tang 경희대학교 2013 국내석사

In a thin client computing architecture, application processing is delegated to a remote server rather than running the application locally. User input is forwarded to the server, and the rendered images are relayed through a dedicated remote display protocol to the user's device. Thin client computing offers the promise of easier-to-maintain computational services with reduced total cost of ownership. The current prevalent thin client protocols nowadays are Citrix MetaFrame, Virtual Network Computing (VNC), Microsoft Remote Desktop (RDP), and Sun Ray. However, the current thin client protocols cannot meet the requirement of modern services, in which most are display-intensive applications, such as videos, games. Based on the surveys in this field, we found that an excellent thin client protocol should be able to satisfy two conditions: 1) Providing efficient data compression ability to reduce network bandwidth consumption; 2) Operating with low computing complexity to decrease response latency and thin client power consumption. Our previous work proposed a solution based on the conditions, in which we invented a novel hybrid remote display protocol (HRDP) for mobile thin client system. VNC-RFB protocol and MJPEG streaming were adopted for low-motion display and high-motion display, respectively. GPU of server was suggested to take a part of JPEG compression task. The motion detection algorithm will determine which area are high motion area that to be handled by MJPEG module, which area are low motion area that to be handled by VNC module. The evaluation results have proved that the protocol provided effective compression, low-latency and guaranteed quality of video. Meanwhile the HRDP still exists some shortages. First, the threshold set in the motion detection algorithm of the HRDP is simply based on system operator's experience. Second, after deployed the thin client system leveraged with HRDP, we found that the server of the system is easy to be crashed due to the exhaustion of server computing resource in the case of multiple users access. Therefore, the remote desktop services will shut down immediately, which give a bad QoE to users. Thus in this thesis, We conducted a series of experiments to explore the QoE model in the HRDP under different user service requirement. The standard error of parameters in the QoE model showed that the estimated function well fitted changes of experiment data. Based on the model, we proposed an adaptive QoE management scheme for feedback with HRDP to solve the above shortages. The simulation results showed us that the thin client system leveraged with the QHRDP provided optimal service by adaptive setting within limited resource.

Trademark issues in entertainment industry

Tang Wei 서울대학교 대학원 2017 국내석사

Entertainment industry is a fast progressing and very profitable business all over the world. Nowadays, there are a lot of cooperation among entertainment enterprises of US, Korea and China. In order to protect and enhance the development of this industry, there is a huge demand for related laws and legal services. However, experts and lawyers specialized in entertainment law always focus on those legal issues most relevant to the regular process of the production, distribution and retailing of an entertainment work, such as agent contract signing and copyright clearing, and sometimes neglect other less relevant issues, like trademark issues. Even in textbooks of entertainment law for US law school students, trademark law is briefly and generally introduced. No specific trademark issue is analyzed thoroughly. However, a lot of cases have already showed the importance of trademark issues in the entertainment industry. As a Chinese trademark law professional who has studied US IP law in a Korean university and who are very interested in entertainment, I am writing this thesis in order to raise the attention on trademark issues in entertainment industry and give suggestions to legislators and lawyers in this area. In the introduction, I explain the importance of trademark issues by analyzing a recent case in China, “the Voice of China” case and listing various advantages of trademark right compared to copyright. In the main body, I divide those important trademark issues into two categories, one before trademark registration and one after that. Before trademark registration, we need to know what to apply, when to apply and who to apply. I introduce some special types of trademarks that are usually used by entertainment enterprises, such as name of celebrities, works and characters, three dimensional marks, sound marks and motion marks. I analyze the requirements of trademark registrability, including priority of the right, distinctiveness, legality and etc. I focus on those special issues related to entertainment industry, like the cross-class characteristic of entertainment service and trademark squatting. In my opinion, nowadays in modern society with high developed technologies, goods may have multi functions and services may have multi ways and purposes. Thus, it is better to be open minded in admitting double characters of one good or service. I analyzed the “星光大道”(Star Road) case and “非诚勿扰”(Feichengwurao) case in details to prove my idea. As to trademark squatting, the most serious problem in China, I illustrate three reasons to it and give respective suggestions to solve the problem. I also give advice to Chinese examiners on quoting a comparatively appropriate article within the range of legality requirement to refuse bad faith applications. Furthermore, I introduce a special requirement on trademark application. That is “not an illegal conversion of copyright and related right.” In the end of this part, I also explain how to make a good deal on the right to apply for trademarks. After trademark registration, we need to know how to use, protect and restrict a registered mark. We shall avoid cancellation of a mark on non-use ground, and prevent infringement, especially dilution and reverse confusion. At the same time we shall respect trademark fair use, particularly the fair use by entertainment enterprises, including descriptive fair use, functional fair use, prior right fair use, nominative fair use and so on. In this part, I especially introduce a new article in Chinese Trademark Law on “no liability for damages to a non-used mark”, mention the difficulty in asserting dilution and suggest entertainment enterprise to avoid reverse confusion. For trademark fair use, I list different definitions and illustrations given by different trademark law systems, and remind entertainment enterprise to refrain from some controversial issues like trademark parody and comparative advertising. From the discussion, we may find out that the trademark protection systems in US, Korea and China related to entertainment industry are in different developing stages. Moreover, due to the complexity of entertainment industry and its related IP rights, it is hard to make a good balancing between trademark right and other IP rights, or a clear distinction between use and misuse, fair use and infringement of trademark right. Hence in conclusion, I suggest legislators and scholars in this area to learn from those advance theories and practices to promote the development of its own country's trademark protection system and entertainment industry. To sum up, there are three main characteristics of my thesis. The first one is about the structure. I arrange those trademark issues in the time flow of a trademark’s designing, application, registration, using and protection, so that readers may have a good understanding about the relations between those trademark issues. The second one is about the comparison. As to almost all issues, I list the similarities or differences among US, Korea and China, so that readers may know in different countries, they need to deal with trademark issues differently. The third characteristic is about examples. I carefully select trademarks and cases closely related to entertainment industry, so that readers may feel the importance of trademark issues to this industry.

Anton Stepanovich Arensky의 <12 Etudes, Op.74> 분석연구

Tang, Wei 세종대학교 대학원 2022 국내박사

국문초록 안톤 스테파노비치 아렌스키(Anton Stepanovich Arensky, 1861-1906)는 19세기 말 러시아의 유명한 작곡가, 피아니스트이자 지휘자이다. 본 연구는 아렌스키 <12 Etudes, Op.74>의 12곡 피아노 작품을 대상으로 이 작품에 담긴 독특한 음악 스타일과 민족성을 탐구하였으며, 이를 통해 아렌스키 피아노 작품의 독특한 음악적 특징을 연구하였다. 현재까지 이 작품에 대한 연구가 거의 진행되지 않아 아렌스키의 <12 Etudes Op.74>에 대한 연구가 부족하다. 따라서 본 연구자는 기존 문헌을 바탕으로 해당 작품에 대한 분석과 연구를 시도하고, 이를 통해 아렌스키의 작곡 스타일을 이해하고자 한다. 본 논문에서는 아렌스키의 <12 Etudes, Op.74>에 대한 연구 필요성과 목적, 연구 범위와 현황에 대해 알아보았고, 음악 분석의 연구 범위과 방법을 제시하였다. 또한 아렌스키의 생애, 작곡 시대 배경과 작품 경향, 그의 피아노 작품들을 알아보고 <12 Etudes Op.74>의 모든 작품을 분석하였다. 아렌스키의 <12 Etudes, Op.74> 작품이 지닌 음악적 특징은 다음과 같다. 첫 번째, 곡의 구조는 3부 형식을 위주로 하고, 정연한 곡의 구성을 지니고 있다. 두 번째, 민족주의과 낭만주의 특징이 뚜렷하여, 음악은 가창적 선율선의 표현을 중요시한다. 또한 서정적인 선율이 전체 곡을 지배한다. 세 번째, 작품 전체에는 다성부의 텍스쳐 구조가 관통하고, 주제 선율이 한 성부에서 다른 성부로 이전되고, 주제 선율이 각 성부에서 교대로 나타난다. 네 번째, 선율 모방 진행, 조성 전환 모방 진행, 축소 모방 진행을 많이 사용하여 주요 음형을 발전시켰다. 다섯 번째, 셋잇단음표와 여섯잇단음표의 리듬 유형을 많이 사용하였다. 본 연구자는 이 논문을 통해 더 많은 연주자들이 아렌스키의 작곡 스타일을 이해하며, 음악적 특징을 발견하고 작품을 연주하는데 많은 도움이 되기를 바란다.

Silicide/Silicon Heterointerfaces, Reaction Kinetics and Ultra-short Channel Devices

Tang, Wei University of California, Los Angeles 2012 해외박사(DDOD)

Nickel silicide is one of the electrical contact materials widely used on very large scale integration (VLSI) of Si devices in microelectronic industry. This is because the silicide/silicon interface can be formed in a highly controlled manner to ensure reproducibility of optimal structural and electrical properties of the metal-Si contacts. These advantages can be inherited to Si nanowire (NW) field-effect transistors (FET) device. Due to the technological importance of nickel silicides, fundamental materials science of nickel silicides formation (Ni-Si reaction), especially in nanoscale, has raised wide interest and stimulate new insights and understandings. In this dissertation, in-situ transmission electron microscopy (TEM) in combination with FET device characterization will be demonstrated as useful tools in nano-device fabrication as well as in gaining insights into the process of nickel silicide formation. The shortest transistor channel length (17 nm) fabricated on a vapor-liquid-solid (VLS) grown silicon nanowire (NW) has been demonstrated by controlled reaction with Ni leads on an in-situ transmission electron microscope (TEM) heating stage at a moderate temperature of 400 ºC. NiSi2 is the leading phase, and the silicide-silicon interface is an atomically sharp type-A interface. At such channel lengths, high maximum on-currents of 890 (microA/microm) and a maximum transconductance of 430 (microS/microm) were obtained, which pushes forward the performance of bottom-up Si NW Schottky barrier field-effect transistors (SB-FETs). Through accurate control over the silicidation reaction, we provide a systematic study of channel length dependent carrier transport in a large number of SB-FETs with channel lengths in the range of (17 nm -- 3.6 microm). Our device results corroborate with our transport simulations and reveal a characteristic type of short channel effects in SB-FETs, both in on- and off-state, which is different from that in conventional MOSFETs, and that limits transport parameter extraction from SB-FETs using the conventional field-effect transconductance measurements. In addition to application of silicide in Si NW devices, the fundamental materials science of Ni-Si reaction is also of interest, and in-situ TEM has been shown to be a useful tool in obtaining dynamical phase transformation information and therefore providing insights into the new phase formation process. By using in-situ TEM techniques, a new gold catalyzed solid-liquid-solid (SLS) silicide phase growth mechanism in Si NWs is observed for the first time, which shows the liquid mediating growth can be also used in synthesis of metallic silicide nanowires. SLS is analogous to the VLS in both being liquid-mediated, but is fundamentally different in terms of nucleation and mass transport. In our SLS growth at 700 ºC, the Ni atoms are supplied from remote Ni particles by interstitial diffusion through Si NW into the pre-existing Au particle at the tip. Upon supersaturation of both Ni and Si in Au, octahedral shape of Ni disilicide phase nucleates in the middle of the Au liquid alloy, which thereafter sweeps through the Si NW and transform Si into NiSi2. Dissolution of Si by Au(Si,Ni) liquid mediating layer and growth of NiSi2 are shown to proceed in different manners. Using in-situ TEM technique, we also have the chance to present direct evidence that Si (111) twin boundaries and Si grain boundaries on Si NW surface can be efficient heterogeneous nucleation site for the silicide growth. By analyzing the nucleation site favorability, unlike other typical FCC materials like Cu or Si, we infer (111) twin defects in NiSi2 may have high interfacial energy. These results may provide valuable insights into the MOSFET source/drain (S/D) contact silicide formation process when defects are either unintentionally formed during the process or intentionally introduced to engineering the strain along the channel.

Finite-state Markov models for correlated fading channels in wireless communications

Tang, Wei University of Pennsylvania 2001 해외박사(DDOD)

Various types of diversity systems are used as countermeasures for fading in wireless communications. In most previous work, performance analyses for diversity systems assumes independent fading channels, and long-term averaging is carried out under perfect interleaving assumption. Performance of systems utilizing an ARQ scheme depends on the dynamics of fading channels, which is not captured in long-term averaging. Bit-level simulation of the performance can be very time consuming. In this dissertation, we design and build finite-state Markov models representing correlated Rayleigh fading channels under the scenarios of spatial diversity and frequency diversity. These models can provide both static and dynamic properties of correlated fading channels, and are very useful for the performance evaluation of diversity systems. We develop and analyze a methodology to partition the equivalent received signal-to-noise ratio of a diversity system into a finite number of states, which correspond to different channel quality, based on level crossing theory and a time duration criterion for each state. Two popular diversity combining schemes i.e., maximal ratio combining and selection combining are considered in our models, and the worst case for the selection combining is considered as well. We present novel analyses and derivations for the level-crossing rate of the received SNR for different combining schemes under correlated fading channels. The physical models of correlated fading channels for spatial diversity and frequency diversity are studied, and provide justification for our level-crossing rate derivations. The efficacy of our models is verified by numerical calculations and computer simulations for different combining schemes and channel conditions. We consider two correlated equal-power fading channels, and extend our models to two correlated non-equal-power fading channels and to <italic>L</italic> (<italic>L</italic> > 2) uncorrelated equal power fading channels. These extensions broaden the applicability of our finite-state Markov models. We demonstrate the usefulness of our Markov models by two examples. First, we obtain the long time average performance for diversity systems with different diversity combining schemes, and the results agree with what have been obtained by other approaches in literature. Second, we carry out performance analysis for diversity systems utilizing a hybrid ARQ scheme with rate-compatible punctured convolutional codes, showing analytical results that agree with simulations.

Enhancing biodegradation after bioremediation of polycyclic aromatic hydrocarbons in soil

Tang, Wei-Chih Cornell University 2003 해외박사(DDOD)

An investigation was conducted to determine possible treatments to increase the biodegradability of residual pyrene by microorganisms in two loam soils after biodegradation and of polycyclic aromatic hydrocarbons (PAHs) in bioremediated soil collected from a former manufactured-gas plant. In the loams, slurrying, repeated inoculation, heating at 75°C for 18 d or 121°C for 18 h, freezing and thawing for one cycle, drying and wetting, methylene chloride soaking, and the addition of 2,2,4,4,6,8,8-heptamethylnonane (HMN), naphthalene, phenanthrene, anthracene, benzoic acid, palmitic acid, ethylenediaminetetraacetic acid disodium salt, and surfactants (Triton X-100 and Alfonic 810-60 ethoxylate) enhanced further biodegradation of residual pyrene remaining after biodegradation. The pyrene concentrations in treated soils were 18 to 81% lower than those in untreated soils. When the final biodegradation was performed using slurried soil, adding nonaqueous-phase liquids inhibited the further biodegradation of pyrene, and the concentrations in treated soil were 14 and 29% higher than those in untreated soil. In the bioremediated soil collected from a former manufactured-gas plant site, inoculation of a soil suspension, heating at 75°C for 9 to 21 d, and adding 2,2,4,4,6,8,8-heptamethylnonane enhanced the biodegradation of some PAHs. The residual concentrations of those PAHs in treated soils were 11 to 70% lower than those in untreated soils. Adding surfactants did not significantly enhance further biodegradation. The data show that several treatments increased the biodegradability of PAHs remaining after an initial biodegradation. These treatments may thus be useful in bioremediation technologies. Some naturally occurring processes, e.g., freezing-and-thawing and drying-and-wetting, also increased the biodegradability of the pyrene remaining in soil. The risk of increased bioavailability of aged and bioremediated toxicants in soil by naturally occurring processes should be further assessed.

Unified Compositional Models for Visual Recognition

Tang, Wei Northwestern University ProQuest Dissertations & T 2019 해외박사(DDOD)

A core problem in many computer vision applications is visual recognition (including object classification, detection and localization). Recent advances in artificial neural networks (aka "deep learning") have significantly pushed forward the state-of-the-art visual recognition performances. However, due to the lack of semantic structure modeling, most current deep learning approaches do not have explicit mechanisms for visual inference and reasoning. As a result, they are unable to explain, interpret, and understand the relations among visual entities. Moreover, since deep learning tries to fit a highly nonlinear function, it is data hungry and can overfit when the data is not big enough.This thesis studies deep and unified computational modeling for visual compositionality. It is a new mechanism to unify semantic structure modeling and deep learning into an effective learning framework for robust visual recognition. Visual compositionality refers to the decomposition of complex visual patterns into hierarchies of simpler ones. It not only embraces much stronger pattern expression powers, but also helps resolve the ambiguities in the smaller and lower-level visual patterns via larger and higher-level visual patterns.We first present a unified framework for compositional pattern modeling, inference and learning. Represented by And-Or graphs (AOGs), it jointly models the compositional structure, parts, features, and composition/sub-configuration relationships. We show that the inference algorithm of the proposed framework is equivalent to a feedforward network. Thus, all the parameters can be learned efficiently via the highly-scalable back-propagation (BP) in an end-to-end fashion. We validate the model via the task of handwritten digit recognition. By visualizing the processes of bottom-up composition and top-down parsing, we show that our model is fully interpretable, being able to learn the hierarchical compositions from visual primitives to visual patterns at increasingly higher levels. We apply this new compositional model to natural scene character recognition and generic object detection. Experimental results have demonstrated its effectiveness. We then introduce a novel deeply learned compositional model for human pose estimation (HPE). It exploits deep neural networks to learn the compositionality of human bodies. This results in a novel network with a hierarchical compositional architecture and bottom-up/top-down inference stages. In addition, we propose a novel bone-based part representation. It not only compactly encodes orientations, scales and shapes of parts, but also avoids their potentially large state spaces. With significantly lower complexities, our approach outperforms state-of-the-art methods on three benchmark datasets. Finally, we study how features can be learned in a compositional fashion. The motivation is that HPE is inherently a homogeneous multi-task learning problem, with the localization of each body part as a different task. Recent HPE approaches universally learn a shared representation for all parts, from which their locations are linearly regressed. However, our statistical analysis indicates not all parts are related to each other. As a result, such a sharing mechanism can lead to negative transfer and deteriorate the performance. To resolve this issue, we first propose a data-driven approach to group related parts based on how much information they share. Then a part-based branching network (PBN) is introduced to learn representations specific to each part group. We further present a multi-stage version of this network to repeatedly refine intermediate features and pose estimates. Ablation experiments indicate learning specific features significantly improves the localization of occluded parts and thus benefits HPE. Our approach also outperforms all state-of-the-art methods on two benchmark datasets, with an outstanding advantage when occlusion occurs.

Internet-scale information monitoring: A continual query approach

Tang, Wei Georgia Institute of Technology 2003 해외박사(DDOD)

Information monitoring systems are publish-subscribe systems that continuously track information changes and notify users (or programs acting on behalf of humans) of relevant updates according to specified thresholds. Internet-scale information monitoring presents a number of new challenges. First, automated change detection is harder when sources are autonomous and updates are performed asynchronously. Second, information source heterogeneity makes the problem of modelling and representing changes harder than ever. Third, efficient and scalable mechanisms are needed to handle a large and growing number of users and thousands or even millions of monitoring triggers fired at multiple sources. In this dissertation, we model users' monitoring requests using continual queries (CQs) and present a suite of efficient and scalable solutions to large scale information monitoring over structured or semistructured data sources. A CQ is a standing query that monitors information sources for interesting events (triggers) and notifies users when new information changes meet specified thresholds. In this dissertation, we first present the system level facilities for building an Internet-scale continual query system, including the design and development of two operational CQ monitoring systems OpenCQ and WebCQ, the engineering issues involved, and our solutions. We then describe a number of research challenges that are specific to large-scale information monitoring and the techniques developed in the context of OpenCQ and WebCQ to address these challenges. Example issues include how to efficiently process large number of continual queries, what mechanisms are effective for building a scalable distributed trigger system that is capable of handling tens of thousands of triggers firing at hundreds of data sources, how to effectively disseminate fresh information to the right users at the right time. We have developed a suite of techniques to optimize the processing of continual queries, including an effective CQ grouping scheme, an auxiliary data structure to support group-based indexing of CQs, and a differential CQ evaluation algorithm (DRA). The third contribution is the design of an experimental evaluation model and testbed to validate the solutions. We have engaged our evaluation using both measurements on real systems (OpenCQ/WebCQ) and simulation-based approach. To our knowledge, the research documented in this dissertation is to date the first one to present a focused study of research and engineering issues in building large-scale information monitoring systems using continual queries.

Effect of wind-driven rain on deterioration of a tall building: Numerical modeling and field measurements

Tang, Wei Carnegie Mellon University 2002 해외박사(DDOD)

Wind-driven rain is an important factor in determining the extent of calcareous stone erosion and the patterns of surface soiling on buildings. In this study, both numerical modeling and field measurements have been employed to investigate the effect of wind-driven rain on building walls. The work has been conducted at the Cathedral of Learning, a 42-story Indiana limestone building on the University of Pittsburgh campus in Pittsburgh, PA. The numerical method utilizes Computational Fluid Dynamics (CFD) techniques and comprises three steps: (a) using the Re-Normalization Group (RNG) k-ϵ model to calculate the airflow field around the building, (b) simulating raindrop trajectories in the flow field by solving the force balance equations, and (c) estimating driving rain fluxes on building walls by combining raindrop trajectory results and meteorological data. To validate the model, wind-driven rain fluxes have been measured at the Cathedral for a 21-month period. Comparison of model estimates and field measurements for 94 rain events show reasonably good agreement, suggesting that the numerical method is effective in predicting driving rain fluxes on building surfaces. Both model and measurement results show that wind-driven rain is strongly influenced by rainfall intensity, wind speed, wind direction, and building geometry. Qualitative comparisons of the observed soiling patterns with the modeled and measured driving rain fluxes show that white eroded areas on the building surfaces generally receive more driving rain than soiled but less eroded areas. The current soiling patterns are most likely due to the non-uniform distribution of wind-driven rain as a result of long-term trends in wind and rain that interact with the building. This work will enhance the understanding of processes involved in the erosion of buildings by providing a quantitative tool for studying the impact of wind-driven rain on building surfaces. This tool will be useful for the selection of treatment programs for damaged buildings and monuments as well as designing new structures that will be less vulnerable to deterioration.