CTChean Khim Toa

Chean Khim Toa

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Chean Khim Toa has a total of 39 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    WiFi-Based Human Activity Recognition and Fall Detection with Taxonomy, Benchmarks, and Future Directions: A Narrative Review

    by Kok Chung Chua, Kai Liang Lew, Chean Khim Toa, Matthew Alexander Paudianto, Ivan Nathanael, Abel Nathanael Hutapea

    WiFi-based human activity recognition (HAR) and fall detection have emerged as promising alternatives to vision- and wearable-based systems for modern assisted living and clinical monitoring for the elderly. These approaches exploit variations in Channel State Information, capturing amplitude and phase fluctuations across subcarriers to characterize how human motion perturbs these patterns through multipath propagation. This narrative review synthesizes 26 peer-reviewed studies published between 2017 and 2025, focusing on systems evaluated in real-world, non-laboratory environments. While recent deep learning models demonstrate near-perfect benchmark accuracy on fine-grained datasets, cross-environment assessments consistently reveal a severe “reality gap.” Specifically, several reviewed studies reported accuracy declines exceeding 20% when deployed in previously unseen buildings, while multi-user interference reduced recognition accuracy by 21–46%. Furthermore, dataset availability and reproducibility are critical considerations; only a limited number of reviewed studies utilized publicly available datasets, and none provided a standardized benchmarking guide-line. To address these gaps, this review presents a comprehensive taxonomy of post-2017 architectures, a comparative synthesis of laboratory versus deployment performance, and a critical analysis of key implementation challenges. We conclude by proposing a set of recommended reporting practices and outlining actionable future research. In doing so, the paper outlines key recommendations for developing adaptive, location-independent models, emphasizing the need for domain-transfer techniques, cross-building validation, and expanding open datasets to enhance generalizability. Received: 31 August 2025 | Revised: 13 March 2026 | Accepted: 25 June 2026 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data sharing is not applicable to this article as no new data were created or analyzed in this study. Author Contribution Statement Kok Chung Chua: Conceptualization, Methodology, Validation, Formal analysis, Writing – original draft, Writing – review & editing, Visualization. Kai Liang Lew: Resources, Writing – original draft, Writing – review & editing, Project administration. Chean Khim Toa: Conceptualization, Validation, Resources, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition. Matthew Alexander Paudianto: Investigation, Data Curation, Writing– original draft, Writing – review & editing. Ivan Nathanael: Investigation, Data curation, Writing – original draft, Writing – review & editing. Abel Nathanael Hutapea: Investigation, Data curation, Writing – original draft, Writing – review & editing.

  • Open Access

    ResProtoNet: A Skeleton-Aware Few-Shot Framework for Yoga Pose Classification

    by Chean Khim Toa, Kai Liang Lew, Sin Pei Ton

    Deep learning models for yoga pose classification traditionally require large, diverse, carefully annotated datasets, which are costly and time-consuming. While open-source yoga datasets are available, collecting and annotating new ones, especially for complex poses, remains a significant challenge. This limitation motivated the use of few-shot learning (FSL) to enable efficient pose classification under limited data conditions. This study proposes ResProtoNet, combining a ResNet-18 feature extractor with a Prototypical Network classifier, evaluated against a supervised baseline using RGB and skeleton-based images. The latter contains joint-based pose encodings extracted via the MediaPipe model. Experiments conducted on five fundamental yoga poses under 1-shot, 3-shot, and 5-shot configurations demonstrated several findings. First, ResProtoNet consistently outperformed the ResNet-18 baseline, achieving 98.2% accuracy in the 3-shot setting, compared to 97.6%. Second, ResNet-18 consistently delivered the strongest baseline performance across both modalities, underscoring its robustness and justifying its role as backbone. A key contribution is a comprehensive stress test comparing RGB and skeleton modalities within an FSL context. Robustness analysis highlighted that skeleton-based input preserved up to 20% higher accuracy under heavy occlusion and reduced accuracy loss by 5.7% under resolution degradation, confirming resilience when visual information degrades. In contrast, RGB retained advantages under light occlusion and higher-resolution inputs, where texture and background remained informative. Overall, applying FSL to a supervised baseline enables reliable pose classification under limited supervision, substantially reducing dependence on large-scale annotated datasets. ResProtoNet demonstrated strong potential for real-world healthcare and exercise monitoring systems, where annotated resources and data quality are constrained. Received: 31 August 2025 | Revised: 8 January 2026 | Accepted: 22 April 2026 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available in Yoga Poses Dataset [Kaggle] at https://www.kaggle.com/datasets/niharika41298/yoga-poses-dataset, in EasyFSL (PyTorch library for few-shot learning) [GitHub] at https://github.com/sicara/easy-few-shot-learning, in Yoga-82 Dataset [Kaggle] at https://www.kaggle.com/datasets/akashrayhan/yoga-82, and in Blazepose_skeletons_Yoga_82 Dataset [Kaggle] at https://www.kaggle.com/datasets/rashiniyasp/blazepose-skeletons-yoga-82. Author Contribution Statement Chean Khim Toa: Conceptualization, Methodology, Validation, Resources, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition. Kai Liang Lew: Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – review & editing, Supervision, Project administration. Sin Pei Ton: Conceptualization, Formal analysis, Resources.

  • A Narrative Review on Big Data and Social Media Behaviour Analysis for Crisis Response in Thailand During COVID-19 and Flooding Events

    by Chean Khim Toa, Kai Liang Lew, Zhi En Tan, Jie Ying Tan, Jia Xuan Yu, Suleiman Aliyu Babale

    Social media platforms have evolved into critical real-time information hubs during crisis events, generating massive user-generated content that presents both opportunities and challenges for emergency management. Thailand's experience with the COVID-19 pandemic and recurring flood disasters provides valuable insights into leveraging social media big data for crisis response in developing countries. This narrative review synthesizes existing literature on technological frameworks, analytical methods, and practical implementations through comprehensive analysis of published studies and documented case studies. By examining distributed computing platforms, natural language processing, sentiment analysis, and geospatial mapping, this review assesses how Thailand has utilised user-generated content for emergency management. The findings reveal both technological progress and persistent systemic constraints. While initiatives such as the Anti-Fake News Centre demonstrate effective misinformation detection within two hours, significant gaps remain in five key areas, including technological infrastructure fragmentation among 48 disaster management agencies, analytical limitations in Thai-language processing, governance framework deficiencies, stakeholder coordination constraints, and digital inclusivity challenges that exclude vulnerable populations. Despite technological implementations, critical barriers include 96% failure rates in monitoring equipment and limited real-time data integration. The analysis provides a systematic examination of implementation gaps spanning technological, analytical, governance, stakeholder coordination, and inclusivity dimensions while identifying strategic opportunities, including enhanced data quality frameworks, cloud-based scalability solutions, and explainable AI integration, to strengthen Thailand's digital crisis management capabilities.

  • Open Access

    Big Data Analytics Technologies and Implementation Challenges for Smart Infrastructure Development in Malaysia: A Narrative Review

    by Chean Khim Toa, Kai Liang Lew, Zhan Yik Jeremy Wong, Kien Horng Low, Tian Ming Khoo

    This paper reviews Big Data Analytics (BDA) applications for traffic management and security monitoring in Malaysian urban infrastructure, examining implementation barriers and scaling requirements for effective adoption. Through an analysis of international case studies from Singapore, São Paulo, and comparable cities, the review identifies four primary barriers preventing the scaling of integrated BDA in Malaysian cities: technological infrastructure gaps, governance fragmentation, regulatory inadequacies, and implementation capacity constraints. Despite recent legislative developments, including the Data Sharing Act 2025 and existing BDA implementations such as Malaysia City Brain, implementation barriers persist due to Malaysia's federal governance structure and infrastructure limitations, which differ significantly from those of successful city-state implementations. This study proposes a contextualised five-priority implementation approach that emphasises expanding existing corridor-based pilots, leveraging federal coordination mechanisms through the Data Sharing Act 2025, regulatory development, strategic partnerships, and performance measurement systems. The analysis reveals that governance coordination and the development of a regulatory framework should guide the scaling of infrastructure investments for effective BDA adoption. These findings provide Malaysian policymakers with a realistic scaling pathway that builds upon existing implementations, working within institutional and resource constraints to develop integrated urban analytics systems for traffic optimisation and security monitoring.

  • Open Access

    Big Data Analytics in Digital Banking Fraud Detection Technologies and Methods

    by Chean Kim Toa, Kai Liang Lew, Kian Yee Jane Yam, Shi Hui Khoo, Kai-Xuan Adam See

    Digital banking fraud has escalated dramatically with the proliferation of online financial services, causing billions in annual losses and threatening the stability of global economic systems. This paper examines the Big Data Analytics (BDA) technologies and methods for real-time fraud detection in digital banking environments. The paper discusses the evolution from traditional rule-based systems to modern distributed computing frameworks, analysing how Apache Hadoop and Spark enable the processing of massive transaction volumes with varying trade-offs between latency and accuracy. Key machine learning approaches are covered, including supervised methods, unsupervised methods, and hybrid architectures that combine both paradigms. The paper identifies critical implementation challenges across technical dimensions, operational aspects, and regulatory requirements. Emerging trends explored include federated learning for privacy-preserving model training, blockchain integration for cross-institutional fraud detection, and edge computing for ultra-low latency inference. The analysis shows that while individual studies report improvements in detection, challenges remain in real-world validation, model interpretability, and cross-institutional generalizability. The paper concludes with practical recommendations for implementing hybrid streaming-batch architectures, embedding explainability mechanisms, and adopting privacy-preserving techniques. This paper provides insight for researchers and practitioners to understand the current capabilities, limitations, and future trends of BDA in enhancing fraud detection in increasingly complex digital banking ecosystems.


  • Comparative Analysis of Augmented Reality and Virtual Reality in Biomedical Engineering Education: Acceptance and Learning Effectiveness

    by Koi‐Qi Lai, Chow‐Khuen Chan, Nur Afiqah Hashim, Chean‐Khim Toa, Che‐Wei Lin, Sim‐Kuan Goh, Lai‐Kuan Tham

    The rapid growth of digital technologies and the shift towards experiential learning have created new opportunities to support biomedical engineering (BME) education, facilitating students in visualization and complex concepts understanding. This study examines the effectiveness and acceptance of digital education tools, specifically augmented reality (AR) and virtual reality (VR) as alternative learning tools for BME students. A convenient sample of 80 BME students (50 women, 30 men, M age = 21.84 ± 0.95 years) was divided into intervention ( n = 43) and control ( n = 37) groups. Intervention participants studied AR and/or VR‐enhanced educational materials while control participants studied the traditional PDF materials and/or 2D videos. Both groups completed quizzes assessing learning effectiveness, with intervention group additionally completed pre‐ and post‐acceptance questionnaires. Results revealed a slight increase in VR acceptance scores, while AR yielded significantly higher post‐scores, reflecting greater usefulness, ease of use, enjoyment, and intention to use AR in the future. In terms of effectiveness, quiz scores revealed no significant difference in performance between both AR ( p = 1.000) and VR studies ( p = 0.294), likely due to the heavy cognitive load and distractions from the multi‐modal user interface and interactions. While AR and VR were not primary educational tools for BME students in Malaysia, AR showed promise as a technology tool for enhancing the teaching and learning pedagogy.

  • A Reproducible Benchmark of AdamW-Augmented Lightweight Models for Trash Classification

    by Chean Khim Toa, Kai Liang Lew, Xin Ming Chee, Shyan Lee Chia

    Global waste generation is projected to reach 3.40 billion tonnes by 2050, creating urgent demands for automated waste classification systems that can overcome the limitations of manual sorting methods. Current deep-learning research on waste classification lacks standardised evaluation protocols, preventing meaningful architectural comparisons and hindering the progress of reproducible research. This paper establishes a reproducible benchmark framework for lightweight neural network models designed explicitly for trash classification research applications. Lightweight models are designed for optmised architecture and computation cost while maintain accuracy. Four representative lightweight models, including MobileNet V3 Large, Vision Transformer (ViT) Small, EfficientFormer, and ShuffleNet V2, were systematically evaluated on the TrashNet dataset using identical training protocols. All models employed AdamW optimisation with a learning rate of 1 × 10-4, weight decay of 1 × 10-4, and CosineAnnealingLR scheduling through 5-fold stratified cross-validation on RTX 2080 Ti hardware. Experimental results demonstrate that ViT Small achieved the highest classification accuracy at 0.815 but required 21.67M parameters, while MobileNet V3 Large delivered superior computational efficiency with 0.768 accuracy and 0.72ms inference time using only 4.21M parameters. Statistical analysis revealed significant performance differences across models (p = 0.0002), with hardware-aware architectural optimisations proving more critical than raw parameter reduction for computational performance on data centre GPU hardware. The standardised evaluation framework and open-source implementation provide rigorous baselines for advancing automated waste classification research.


  • A Narrative Review of Data Mesh Architecture Principles and Implementation Outcomes

    by Kai Liang Lew, Chean Khim Toa, Cheng Hong Chew, Xi Yuan Wong, Suleiman Aliyu Babale

    Centralised data architectures often create operational bottlenecks that limit organisational agility. Data Mesh offers a distributed alternative through domain ownership and federated governance. This narrative review synthesises 52 sources published between 2001 and 2024, examining the evolution from traditional data architectures to Data Mesh implementations across financial services, healthcare, e-commerce, and technology sectors. The review traces the progression from centralised data warehouses through distributed computing frameworks to Data Mesh's emergence, identifying four foundational principles domain-oriented decentralisation, data as a product, self-serve infrastructure, and federated governance. Analysis of recent implementation studies reveals mixed outcomes. Successful adoptions demonstrate improved domain autonomy and reduced central bottlenecks. However, multiple case reports significant coordination complexity and extended implementation timelines, with transformations requiring substantial investments in platform engineering. Consistent challenges emerge, including skill gaps in domain teams transitioning to data ownership, policy conflicts in federated governance structures, infrastructure investments that exceed traditional architectures, and cultural resistance to distributed accountability. Implementation success correlates with existing DevOps maturity, sustained executive sponsorship, phased adoption approaches, and robust metadata management capabilities. The review identifies critical research gaps in standardised success metrics, quantitative failure analysis, privacy-preserving techniques for federated environments, and long-term sustainability assessment. Based on the analysed cases, Data Mesh appears most suitable for large enterprises with diverse data domains and established platform engineering capabilities. Smaller organisations may find centralised approaches more appropriate given the complexity and resource requirements of distributed architectures. This synthesis provides practitioners with evidence-based insights while highlighting priorities for future research.

  • Open Access

    AI-Assisted Analysis for Breast Cancer Imaging and Diagnostics

    by Kai Liang Lew, Chean Khim Toa, Pengfei Zhou, Chia Shyan Lee, Tetuko Kurniawan, Suleiman Aliyu Babale, Cheng Zheng

    Breast cancer cases have increased by 0.5% each year. X-ray, CT-Scan, and magnetic resonance imaging have been used to detect cancer without harming the patient. However, these methods usually used manual screening to process medical images, which leads to longer processing time and increases the burden on the expert. With the help of deep learning, automation-driven breast cancer detection, segmentation, and explanation can be performed in the process, which can greatly reduce the processing time and the burden on experts. This paper proposed a deep learning model, S-YOLOv11 by combining YOLOv11 with SimAM attention mechanism and a GUI with integration of a large language model. The model is trained with 624 images and tested with 156 images. Several YOLO architectures were compared, including YOLOv8, YOLOv9, YOLOv10, and YOLOv11. The proposed model has outperformed the other models. In the detection task, 0.806 precision, 0.635 recall, and 0.724 mAP were achieved. In the segmentation task, 0.833 precision, 0.65 recall, and 0.739 mAP were achieved. In addition, the study also improved the functionality of the GUI by accessing the ChatGPT API. It is possible to generate medical analysis for breast cancer tumors, with the use of GUI for visualization. However, current research is still in the development stage and it needs to be put into clinical trials before it can be used.


  • Open Access

    Development of Virtual Reality-Based Left Brain System

    by Kok Swee Sim, Kok An Teo, Chean Khim Toa, Kai Liang Lew

    This paper proposes a virtual reality based Left Brain System named BrainUp World to improve left brain thinking. The left-brain system was employed with mobile virtual reality technology, hand motion tracking and haptic feedback system. The implementation of these systems is to enhance the experience and sense of embodiment. BrainUp World includes virtual reality-based brain training games to improve a person’s attention level, reasoning skill, auditory cognition, arithmetic skill, sequencing skill and memory. The hand tracking system was utilized with an IR camera to capture the hand orientation and return the gesture data to the VR application. A low-cost and lightweight haptic glove was invented which provides the sense of touch using vibrations while interacting with VR contents. An experimental study was conducted to assess the efficacy of BrainUp World compared to traditional PC-based training approaches. Participants were randomly assigned to either the VR-based or PC-based training group and underwent 6 different games to test a person’s attention level, reasoning skill, auditory cognition, arithmetic skill, sequencing skill, and memory. The results revealed a statistically significant improvement in left brain function in the VR-based training group compared to the PC-based group, with a T-score growth of 4.73. Analysis using ANOVA confirmed the significance of this difference (p-Value = 0.04). Notably, the study also identified age-related differences in thinking fluency, highlighting the importance of personalized cognitive training interventions. In conclusion, BrainUp World demonstrates the potential of VR technology in promoting left brain development, as evidenced by empirical findings from the conducted study. By offering immersive and interactive cognitive training experiences, VR-based systems hold promise for enhancing various aspects of cognitive function associated with left hemisphere dominance. Further research in this area is encouraged to explore the full potential of VR-based interventions for cognitive enhancement.