13474 research outputs found
Sort by
A deep learning approach for facial detection in targeted billboard advertising / Lau Sian En
No full textDeep learning has significantly changed industries by facilitating the development of more intelligent and adaptive systems, with applications especially in advertising. Facial detection using deep learning in advertising offers the potential for highly personalised and effective marketing by leveraging real-time consumer demographics. This research explores a deep learning-based facial detection system for targeted advertising, aiming to enhance consumer engagement by delivering personalized advertisements. The research focuses on addressing key difficulties including lack of audience-targeted delivery, real-time implementation challenges and model accuracy difficulties. This system utilises sophisticated deep learning algorithm using Convolutional Neural Network (CNN) to identify and examine human faces, enabling advertisers to customise their content according to demographic variables including age and gender. The system has two modules which are the Realtime Module and the Dataset Evaluation Module. The system employs Multi-task Cascaded Convolutional Networks (MTCNN) for face detection in the DeepFace model, processes webcam photos, predicts age and gender, and maps relevant advertisements accordingly. The evaluation process encompasses real-time performance analysis and testing using the Wikipedia dataset, evaluating the accuracy, precision, recall, F1-score, and confusion matrices. The system’s capacity to provide targeted advertising not only enhances user experience but also greatly enhances consumer engagement. Results indicate that the Realtime Module attains an accuracy of 70% in age prediction and 90% in gender prediction, whereas the Dataset Evaluation Module achieves an accuracy of 74% for age prediction and 90% for gender prediction, hence enhancing advertisement relevance. The study indicates that using facial recognition technologies in advertising tactics can transform conventional advertising methods, providing real-time, adaptive solutions customised for diverse audiences
Integration of visual art in learning figurative language among EFL English literature undergraduate students / Clara Ling Boon Ing
No full textThis research explores the integration of visual arts as a promising area for scholarly research, offering creative approaches that reshape human experience across diverse contexts. The study investigates the potential of incorporating visual arts to enhance the impacts of learning within the realm of literature education. Previous research has shown that students learning English as a Foreign Language (EFL) often face challenges with figurative language, making literary texts difficult to comprehend. To address this, the study employs a systematic Arts-Based Action Research (ABAR) methodology to explore how drawing can help undergraduate EFL students grasp figurative language concepts. Eleven undergraduate EFL students studying English Literature were selected as participants for the main study. This qualitative action research, conducted over three and a half months, examined innovative strategies for incorporating drawing into literature instruction, going beyond the limitations of traditional text-based learning. Through three research cycles, the researcher refined techniques that integrated drawing with cognitive expression, symbolic art, collaborative learning, and textual connections rooted in cultural contexts. Each cycle involved targeted interventions using visual art elements, design principles, visual thinking strategies, and literature learning concepts. Anchored in Richard Mayer’s Cognitive Theory of Multimedia Learning (CTML), the study revealed that by integrating these strategies, participants engaged more deeply with sensory memory, allowing them to better connect with figurative language by drawing on their cultural backgrounds. This study proposed a way to design interventions to integrate visual art – specifically drawing – not simply as an expressive output, but as a structured, sensory-driven and cognitive learning tool that enhances EFL learners’ understanding of figurative language. The findings also demonstrated improvements in the researcher’s teaching practice, emphasizing the importance of focusing on the learning process rather than the outcome. This study makes a significant methodological contribution to the ABAR framework by providing further evidence of its effectiveness in non-Western settings. Additionally, it extends the application of CTML theory, highlighting the benefits of prolonged sensory memory engagement. The research offers practical insights for educators seeking to integrate visual arts into their teaching practices
A framework for child-friendly urban green spaces in Shanghai: Insights from children’s perspectives / Niu Yunlong
No full textChildren’s activities in urban green space (UGS) are their primary means of connecting with nature and a crucial way to form lasting environmental bonds and promote their physical and mental development. Unfortunately, adults often overlook the importance of UGS in encouraging children’s outdoor activities, resulting in many such spaces’ planning and design failing to truly reflect considerations of children’s needs and desires. Against this backdrop, this study explores the various factors affecting children’s behavior patterns and affordances in UGS and attempts to reveal how these factors correlate with children’s expectations for an ideal child-friendly UGS.
This study was conducted across three UGS (Xujiahui Park, Guilin Garden, and Houtan Riverside) in Shanghai, China. The research framework employs phenomenological methodology, utilizing multiple data collection instruments, including go-along interviews (n=30), behavioral mapping observations (n=30), and participatory drawings (n=108), to examine children's perceptual experiences and behavioral patterns within these environments. The data were analyzed using descriptive statistics and content analysis methods. The findings reveal the diversity and complexity of children’s behaviors in UGS. Children’s activity choices and site usage preferences are significantly influenced by the green space’s (1) Accessibility, (2) Aesthetics, (3) Activities, (4) Safety, and (5) Amenities. Specifically, children prefer UGS with abundant vegetation, safe facilities, and exciting play equipment. Additionally, the study found that children have clear visual and functional expectations for an ideal child-friendly UGS. They wish these places to meet their play and exercise needs and provide comfortable spaces for rest and socialization. These insights are of significant importance to urban planners, emphasizing the need to comprehensively design UGS from a child’s perspective, considering their specific needs and activity preferences. A more humane, inclusive, and child-friendly UGS can be created by achieving this aim
Prioritising the maintenance of public park features and facilities to improve the park user experience / Gan Xing Ni
No full textPublic parks are one of the spaces for all people to do conduct active or passive recreational activity. Features and facilities are provided by authority to support the recreational activity. However, assertions are found that there is problem on the provision and maintenance of features and facilities in public park. Accompanied by the rising maintenance cost, it indicated that there is a need to study the perception of park user on which features and facilities are crucial to provide and maintain. This research ranked on the importance on provision and experience from park user perspective, examine relationship between the maintenance quality and public park performance, and produce a maintenance priority list to facilitate authority in managing more effectively and produce a healthy society. The study adopted mixed-method approach which include quantitative and qualitative methods. In first phase, questionnaires containing perception towards importance on provision, experience in using the variables, public park performance, and recommendation on improvement were distributed to respondent who attended to any public park in Malaysia and a total of 1658 valid questionnaire were collected. The research identified importance of provision and experience using ranking analysis. Subsequently, correlation analysis was performed to test on the performance and a maintenance priority list is constructed. In order to validate the survey result, thirty (30) park goers are chosen for semi-structure interview to acquire further details on importance on provision and maintenance aspect. The findings highlighted ten (10) features and facilities were important to be provided in a park, these are: rubbish bin, washroom, lighting, track or path, signage, natural landscape, park furniture, soft designed landscape, prayer room, and hard designed landscape. In terms of experience, ten (10) features and facilities given user a good experience and nine (9) features and facilities given user a moderate experience. Besides, it is found that the maintenance quality of all features and facilities is significant to public park performance. In the following analysis, ten (10) features and facilities arranged in descending order of importance, prayer room, washroom, signage, hard designed landscape, park lighting, park furniture, rubbish bin, soft design landscape, natural landscape, and track and path were found to be in need to be prioritised in terms of its maintenance
Defects identification on semiconductor wafer for yield improvement using machine learning / Pedram Tabatabaeemoshiri
No full textThe semiconductor industry underpins modern technology, with its products
embedded in almost every electronic device. As semiconductor devices grow increasingly
intricate, ensuring their quality and reliability becomes more challenging. Electrical
testing is crucial to semiconductor wafer quality assurance, designed primarily to identify
fabrication defects. However, the testing process itself can inadvertently introduce new
defects that may go undetected by subsequent inspection methods such as manual and
visual inspection. When these defects escape detection, defective wafers may reach
customers, leading to rejection and return to the manufacturer, resulting in significant
yield losses and operational inefficiencies. This study addresses the urgent issue of
detecting hidden defects in semiconductor wafers that conventional methods overlook.
This work presents a novel graph-based semi-supervised learning (GSSL) algorithm
designed for wafer defect detection. The proposed methodology involves collecting wafer
inspection data, extracting relevant features, and applying the GSSL algorithm to identify
the hidden defects. The approach constructs a graph representation of the wafer,
leveraging its physical layout and test configuration, and integrates domain-specific
knowledge. The method uses weighted edges to represent the likelihood of defect
propagation between dies, optimized through extensive experimentation, followed by an
iterative label propagation process to uncover hidden defects. Experimental results
demonstrate the effectiveness of our method, achieving a 68% accuracy in detecting
hidden defects across multiple product categories in real-world semiconductor
manufacturing environments. The algorithm showed consistent performance across
different wafer types and test configurations, outperforming traditional detection methods with improved computational efficiency. This study offers valuable insights into the
semiconductor industry, providing an advanced tool to enhance yield management and
quality control processes
Joint optimization of resources allocation for quality of service aware next-generation heterogeneous cellular networks / Hayder Faeq Rasool Alhashimi
No full textThe next generation mobile communication system is expected to meet the different service needs of modern communication scenarios. Heterogeneous networks (HetNets) have received a lot of attention in recent years due to their potential as a novel paradigm for evolutionary networks. When compared to homogeneous networks, HetNets provide more potential for spatial spectrum reuse and higher Quality of Service (QoS). However, effective resource management solutions are essential to reduce interference and accomplish spectrum sharing. The research seeks to identify key challenges with the goal of developing effective approaches in 6G HetNets. In this context, resource management aspects such as user association, spectrum allocation, and power allocation are studied. The Millimeter wave band (100 GHz) for HetNet in the downlink scenario is considered. The first scheme in this thesis, joint user association (UA) and channel allocation (CA) problem in two-tier HetNets to improve QoS is investigated. An innovative scheme for user association and channel allocation is presented, wherein the user can be connected to either Macro Base Station (MBS) or a possible Small Base Station (SBS) in a direct or relay-assisted link. A matching game-based user association is proposed to find the optimal association for users. Moreover, a modified auction game is applied to allocate the optimal channel by considering the quota of each next-Generation Node Base Station (gNB). The second scheme in this thesis, a joint user association, channel assignment, and power allocation optimization scheme in 6G HetNets is proposed. The proposed optimization problem is divided into three subproblems. First, a greedy-based user association algorithm is proposed to allocate multiple users to MBS or SBSs to maximize the total sum rate. Then, an auction algorithm is proposed to solve the channel assignment optimization problem to maximize the spectrum efficiency (SE). The utility function of the auction algorithm is formulated based on a Multiple Attributes Decision-Making (MADM) strategy in which the considered attributes are SE and Energy Efficiency (EE). Finally, a State-Action-Reward-State-Action (SARSA) algorithm, which is a reinforcement learning approach, is proposed to solve the power allocation optimization problem. The SARSA algorithm determines optimal power allocation that enhances EE by investigating different power levels. The simulation results of the first scheme showed that the proposed joint UA and CA approach performs well over the state-of-the-art techniques in terms of connection probability, throughput, EE, and SE. Moreover, the significance of evaluating the MBS power level is demonstrated, in terms of probability of connection, and average data rate. The auctions allow for load balancing between the macro channels allocated to SBSs. The effectiveness of the proposed scheme in a relay-assisted scenario is demonstrated in terms of data rate, SE, EE, outage probability, and total saving power. On the other hand, the simulation results of the second scheme illustrated that the proposed SARSA algorithm outperforms the benchmark reward functions, random scheme, and maximum power-based framework in terms of network performance. Furthermore, the proposed approach demonstrates tradeoffs between SE and EE. The outcomes of this thesis demonstrate effective resource utilization that minimizes losses while achieving outstanding results with the same resources
Composite of graphene oxide impregnated palm kernel shell based activated carbon for dye wastewater treatment / Tan Yan Ying
No full textDye wastewater generated from various industrial processes can lead to substantial pollution and environmental damage due to its persistent and recalcitrant nature. Among the various wastewater treatment technologies, adsorption using green and low-cost biomass-based materials stands out as a promising approach for treating dye wastewater. However, raw biomass-based adsorbents such as palm kernel shell (PKS), rubber seed shell, and coconut shell often have limitations in selectivity, recovery, and adsorption capacity, which restrict their use in treating different types of wastewater. This study aims to enhance the adsorption performance of raw PKS activated carbon (PKSAC) through the impregnation with graphene oxide (GO) and iron oxide. Physicochemical analyses revealed that after impregnation with GO and iron oxide, the ternary composite exhibited a significant increase in oxygen content (from 6.01% to 49.73%), a larger pore width (from 3.00 nm to 9.99 nm), and increased oxygenated functional groups compared to raw PKSAC. The adsorption performance of PKSAC-based adsorbents was evaluated for treating synthetic wastewater containing an anionic dye. The adsorption study demonstrated that the ternary composite performed better, achieving 99.8% color removal efficiency and an adsorption capacity of 27.3 mg/g, compared to other PKSAC-based adsorbents (60.7%-73.1%; 16.6 mg/g-19.9 mg/g) under optimum conditions. The ternary composite showed superior performance in chemical oxygen demand and color removal efficiencies, surpassing commercial activated carbons (CACs) by 28.4% and 31.9%, respectively. The reusability of the composite was confirmed over five cycles, maintaining 74.1% performance, significantly higher than CACs (30.0%). Furthermore, the iron leaching (<0.3 mg/L) was negligible for drinking water, confirming the stability and safety of the composite. To further improve adsorption performance, the synthesis process of the ternary composite was optimized. This optimized ternary composite demonstrated an adsorption capacity of 76.4 mg/g for actual industrial printing wastewater, which was 34.5% higher than the composite before optimization (56.8 mg/g). Experimental results indicated that adsorption using the ternary composite is predominantly a monolayer chemisorption process. A quantum chemical analysis revealed adsorption energies below -50 kJ/mol for all functional groups. Feature importance analysis using machine learning revealed that the chemical properties of the adsorbents had a more significant impact on adsorption performance than physical properties. This study also evaluated the possibility of using the spent adsorbent in brick formation, in addition to conducting leaching and phytotoxicity studies. In summary, the research proved that the ternary composite developed from PKSAC is an efficient
adsorbent for treating dye wastewater, offering enhanced performance, stability, and practicality for industrial applications
Extraction of heavy metals and phenolic pollutants from environmental systems utilizing designer green solvents / Irfan Wazeer
No full textWater pollution is a critical and problematic issue that threatens the sustainability of human civilization. Heavy metals such as cadmium and lead are responsible for myriad environmental problems due to their toxicity. In addition, phenol and cresol isomers are classified as priority phenolic pollutants due to their high toxicity to human health and aquatic life. Therefore, these pollutants must be removed from waste streams before they are released into the environment. Several volatile organic compounds (VOCs) that have been used as extractants are toxic, volatile, and flammable. A class of neoteric solvents called hydrophobic deep eutectic solvents (HDES) have recently attracted considerable interest from both academia and industry. HDES are generally immiscible with water solutions and exhibit high extraction efficiency for various target analytes. The objective of this thesis is to investigate the feasibility of HDES for the removal of heavy metals and phenolic contaminants by liquid-liquid extraction (LLE) processes. Conductor-like Screening Model for Real Solvents (COSMO-RS) was used for the possible selection of potential HDES. In addition, different correlations were used to ascertain the reliability of the experimental data. Based on the COSMO-RS screening and the availability of chemicals in the laboratory, some potential HDES were selected for the extraction of phenolic contaminants. The HDES were characterized by measuring their main physical properties such as melting point, stability, viscosity, and density. To understand the formation of intermolecular interactions such as hydrogen bonds between the precursors of HDES, FTIR and 1HNMR analyses were performed. For the removal of cresols, six HDES were experimentally investigated, and all HDES showed very high efficiency in removing cresols from water. The effects of contact time, mass ratio of HDES to water, initial concentration, and molar ratio of HDES were also investigated for the three selected HDES. The extraction efficiency of > 94% was achieved for the removal of cresol isomers from wastewater with all prepared HDES. For the removal of phenol, the TOPO-based HDES showed higher extraction efficiency (up to 96%). The study also examines the extraction of lead and cadmium with eight HDES. Among eight HDES, thymol:decanoic acid (1:1 molar ratio) showed the highest efficiency: 93.49% for lead at 1000 ppm and 76.70% for cadmium at 100 ppm. Optimization of parameters such as HDES molar ratio, contact time, pH and HDES to water mass ratio further improves performance. Regeneration and reuse of the HDESs has proven effective over multiple cycles, with minimal loss of efficiency. Terpene-based HDESs have also been investigated for the extraction of iron and copper. Thymol:decanoic acid shows an extraction efficiency of 93.91% for iron at 100 ppm, while menthol:decanoic acid achieves an efficiency of 74.69% for copper at 10 ppm. The extraction mechanism is explored using FTIR spectra and the solvents show high reusability and sustainability. In this study, a total of 10 HDES are utilized. The results highlight the effectiveness of HDESs as sustainable and scalable solutions for environmental remediation
Predictive dynamic CFD approach to reducing airborne transmission in naturally ventilated hospital rooms: Impact of window opening angles during transitional cold seasons in China / Haowei Yu
No full textThe SARS-CoV-2 airborne virus outbreak has once again drawn attention from
researchers to the features of viral cross-transmission, particularly the numerous
cross-transmissions that took place in hospitals. Autodesk Computational Fluid
Dynamics (CFD) is the most widely used simulation software; nevertheless, technical
limitations hinder it from effectively reproducing the dispersion and transmission
properties of viruses in naturally ventilated rooms under actual conditions. As a result,
there is no established way to prevent the spread of viruses across rooms rely only on
natural airflow currently.
In this thesis, a CFD simulation approach is proposed to investigate the diffusion
characteristics of airborne viruses/pollutants in multiple rooms on the same level,
considering different window statuses (opening angles). This novel simulation approach
combines a window state prediction algorithm and CFD simulation technology.
The first approach predicts the window angle based on indoor and outdoor
environmental factors obtained from fieldwork measurements. Stepwise polynomial
regression has been validated as a novel algorithm that can effectively predict window
opening angles. The second approach develops a dynamic CFD model for the target
building. The UDF was used to compile dynamic boundary conditions. By combining
these two models, more realistic and dynamic characteristics of indoor pollutant dispersion are simulated and extracted.
The study found that using this novel CFD simulation framework allows for a relatively
accurate description of velocity and concentration fields in naturally ventilated buildings.
Specifically, the simulation accuracy for the velocity field is approximately 75%. The
accuracy for the concentration field is even higher, with a maximum APE of only 25%
and an average prediction accuracy of 97.2%. Furthermore, the study found that if the
window opening direction is improperly set, window-opening behavior may lead to
severe pollutant dispersion rather than effectively reducing indoor pollutant levels. This
study not only clarified the practical application of the window-opening behavior
prediction model but also provided a valuable reference for future dynamic CFD
simulation studies on buildings with central corridors
A new cross-overlapped decoupling coil structure for EV dynamic inductive wireless charging system / Yin Jia Lin
No full textThe multi-transmitter and multi-receiver structure has been extensively discussed in Dynamic Inductive Wireless Charging (DIWC) systems for Electric Vehicles. Multiple coils have already been proven successful in high-power applications. Nevertheless, challenges persist in mitigating the adverse effects of cross-coupling among adjacent multi-coils and achieving higher transfer efficiency and stable output power under dynamic conditions. Thus, this paper proposes a new DIWC magnetic coupler consisting of a cross-overlapped transmitter and single receiver that serves multiple purposes, including decoupling, reduced mutual inductance fluctuation, and achieving high efficiency simultaneously. Moreover, the traversal method uses the finite element analysis tool ANSYS Maxwell to choose the optimized Rx coil turns and size, further improving system performance. A 2kW wireless charging setup is developed to validate the proposed DIWC magnetic coupler. The experimental results show that the output voltage fluctuation can be controlled within ±1%, and the maximum DC-DC efficiency is 92.87%. The longitudinal misalignment performance of the proposed DIWC magnetic coupler is also elaborated in this paper