PCPimpa Cheewaprakobkit

Pimpa Cheewaprakobkit

*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


Pimpa Cheewaprakobkit has a total of 20 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Fake news detection through multi-level contextual and sequential modeling

    by Pimpa Cheewaprakobkit, Rindra Razafinjatovo

    The spread of false information on digital platforms destabilizes democratic societies and undermines public trust. Conventional machine-learning models often fail to capture the subtle contextual cues and long‑range linguistic dependencies required for accurate verification. To address these limitations, we propose a hybrid deep‑learning architecture that integrates three modules: A Contextual Encoder Module (CEM) using RoBERTa to generate relational embeddings for deep semantic extraction; a Bidirectional Sequence Learner (BSL) to model long‑range temporal dependencies; and a Context Refinement Layer (CRL) that employs attention mechanisms to highlight the most salient deceptive markers. We evaluated the model on the WELFake dataset, which contains 72,134 news articles. Our proposed method achieved 99% accuracy, precision, recall, and F1‑score, significantly outperforming state‑of‑the‑art baselines, including BERT (95%) and BiLSTM (97%). The CRL's attention outputs provide transparency into the model's decision‑making process, an essential feature for applications in misinformation detection and automated content moderation. Beyond its technical contributions, this research supports society by enabling early and reliable detection of false information, reducing harm from misinformation, and promoting a safer, more trustworthy digital environment.

  • Melodic Skeleton Generation Using Simulated Annealing

    by Yu-Cheng Lin, Cheng-Han Wu, Pimpa Cheewaprakobkit, Yu-Wen Lei, Timothy K. Shih


  • Open Access

    Automatic Timbre Transformation Using Enhanced Diffusion Model

    by Cheng-Han Wu, Pimpa Cheewaprakobkit, Timothy K. Shih, Yu-Cheng Lin, Bing-Ze Liu

    We present a novel timbre transfer model that uses an enhanced diffusion architecture to convert music from various instruments into Erhu timbre. The Erhu, a traditional Chinese instrument, is difficult to simulate due to its rich vibrato and smooth note transitions. Existing Musical Instrument Digital Interface systems struggle to capture its nuanced dynamics. Our model integrates a Pitch Encoder, a Loudness Encoder, and a Diffusion Decoder. The encoders extract pitch features and dynamic loudness variations, guiding the decoder in generating realistic Erhu timbre. By extracting general musical features, the system generalizes to unseen input types without retraining. Evaluations based on pitch accuracy, cosine similarity, and Fréchet Audio Distance show that our model achieves 96% pitch accuracy and high fidelity in Erhu timbre reproduction. This study demonstrates the potential of diffusion-based timbre transfer models in music generation and provides new directions for future work on both music generation and timbre transfer.

  • Open Access

    Examining the Factors Influencing Employee Retention and Loyalty in a Private Hospital in Bangkok, Thailand

    by Damrong Sattayawaksakul, Pimpa Cheewaprakobkit, Busabong Chulapetch

    Employee retention and loyalty are critical for operational efficiency and quality patient care in hospitals. This study investigates the key factors influencing these outcomes through Structural Equation Modeling (SEM) using survey data from 332 employees of a private hospital in Bangkok. The model examined compensation, dedication, engagement, leadership, teamwork, and career advancement. Results revealed dedication (β = 0.62, p < 0.01) and teamwork (β = 0.83, p < 0.01) as the strongest positive predictors of retention and loyalty, while career advancement showed a significant negative direct effect (β = –0.66, p < 0.05) when not supported by internal opportunities. Compensation positively influenced retention (β = 0.27, p < 0.01), and leadership, though beneficial for teamwork and career growth, had a small negative direct effect (β = –0.05). These findings suggest strategic actions such as enhancing leadership development, fostering collaborative cultures, aligning career pathways with internal opportunities, and offering competitive compensation packages. By implementing these targeted strategies, hospital administrators can reduce turnover, strengthen workforce commitment, and improve healthcare service delivery.

  • Open Access

    Enhancing Multi-Object Tracking with Compact Model Adjustment

    by Pimpa Cheewaprakobkit

    Tracking human movement and interactions in complex environments is a key challenge in computer vision, especially for multi-object tracking. Transformer-based models have shown promise in addressing these challenges due to their capacity to recognize complex patterns across sequences. However, their high computational demands and substantial training data requirements often restrict their real-world applicability. This study aimed to enhance multi-object tracking by introducing a Compact Model Adjustment approach that integrates trainable rank-decomposition matrices within the Transformer architecture. This approach involves freezing the pre-trained model weights and adding trainable low-rank matrices to each layer, substantially reducing the number of parameters that need updating during training. This design allows the model to retain its pre-trained knowledge while efficiently adapting to new tasks, thereby reducing the overall computational load. Additionally, the proposed approach utilizes data from both the current and previous frames to refine object localization and association. Experimental results on the MOT17 benchmark demonstrated that this method achieved a Multiple Object Tracking Accuracy of 71.0, comparable to state-of-the-art techniques while enhancing computational efficiency. This work provides a practical solution for real-world applications in areas such as surveillance, autonomous driving, and sports analytics.

  • Open Access

    Using Hybrid Models for Action Correction in Instrument Learning Based on AI

    by Avirmed Enkhbat, Timothy K. Shih, Munkhjargal Gochoo, Pimpa Cheewaprakobkit, Wisnu Aditya, Thai Duy Quy, Hsinchih Lin, Yu-Ting Lin

    Human action recognition has recently attracted much attention in computer vision research. Its applications are widely found in video surveillance, human-computer interaction, entertainment, and autonomous driving. In this study, we developed a system for evaluating online music performances. This system conducts experiments to assess performance of playing the erhu, the most popular traditional stringed instrument in East Asia. Mastering the erhu poses a challenge, as players often struggle to enhance their skills due to incorrect techniques and a lack of guidance, resulting in limited progress. To address this issue, we propose hybrid models based on graph convolutional networks (GCN) and temporal convolutional networks (TCN) for action recognition to capture spatial relationships between different joints or keypoints in a human skeleton, and interactions between these joints. This can assist players in identifying errors while playing the instrument. In our research, we use RGB video as input, segmenting it into individual frames. For each frame, we extract keypoints, encompassing both image and keypoint information, which serve as input data for our model. Leveraging our innovative model architecture, we achieve an impressive accuracy rate exceeding 97% across various classes of hand error modules, thus providing valuable insights into the assessment of musical performances and demonstrates the potential of AI-based solutions to enhance the learning and correction of complex human actions in interactive learning environments.


  • Open Access

    Human Action Recognition and Note Recognition: A Deep Learning Approach Using STA-GCN

    by Avirmed Enkhbat, Timothy K. Shih, Pimpa Cheewaprakobkit

    Human action recognition (HAR) is growing in machine learning with a wide range of applications. One challenging aspect of HAR is recognizing human actions while playing music, further complicated by the need to recognize the musical notes being played. This paper proposes a deep learning-based method for simultaneous HAR and musical note recognition in music performances. We conducted experiments on Morin khuur performances, a traditional Mongolian instrument. The proposed method consists of two stages. First, we created a new dataset of Morin khuur performances. We used motion capture systems and depth sensors to collect data that includes hand keypoints, instrument segmentation information, and detailed movement information. We then analyzed RGB images, depth images, and motion data to determine which type of data provides the most valuable features for recognizing actions and notes in music performances. The second stage utilizes a Spatial Temporal Attention Graph Convolutional Network (STA-GCN) to recognize musical notes as continuous gestures. The STA-GCN model is designed to learn the relationships between hand keypoints and instrument segmentation information, which are crucial for accurate recognition. Evaluation on our dataset demonstrates that our model outperforms the traditional ST-GCN model, achieving an accuracy of 81.4%.


  • Open Access

    Rule-based explaining module: Enhancing the interpretability of recurrent relational network in Sudoku solving

    by Pimpa Cheewaprakobkit, Timothy K. Shih, Timothy Lau, Yu-Cheng Lin, Chih-Yang Lin

    Computer vision has gained significant attention in the field of information technology due to its widespread application that addresses real-world challenges, surpassing human intelligence in tasks such as image recognition, classification, natural language processing, and even game playing. Sudoku, a challenging puzzle that has captivated many people, exhibits a complexity that has attracted researchers to leverage deep learning techniques for its solution. However, the reliance on black-box neural networks has raised concerns about transparency and explainability. In response to this challenge, we present the Rule-based Explaining Module (REM), which is designed to provide explanations of the decision-making processes using Recurrent Relational Networks (RRN). Our proposed methodology is to bridge the gap between complex RRN models and human understanding by unveiling the specific rules applied by the model at each stage of the Sudoku puzzle solving process. Evaluating REM on the Minimum Sudoku dataset, we achieved an accuracy of over 98.00%.


  • Open Access

    Enhancing Single Object Tracking With a Hybrid Approach: Temporal Convolutional Networks, Attention Mechanisms, and Spatial–Temporal Memory

    by Pimpa Cheewaprakobkit, Chih-Yang Lin, Timothy K. Shih, Avirmed Enkhbat

    Deep neural network-based tracking tasks have experienced significant advancements in recent years. However, these networks continue to face challenges in effectively adapting to appearance changes in both target and background, as well as linking objects after extended periods. The primary challenge in tracking lies in the frequent changes in a target’s appearance throughout the tracking process, which can potentially reduce tracker robustness when faced with issues such as aspect ratio changes, occlusions, scale variations, and confusion from similar objects. To address this challenge, we propose a tracking architecture that combines a temporal convolutional network (TCN) and attention mechanism with spatial-temporal memory. The TCN component empowers the model to capture temporal dependencies, while the attention mechanism reduces computational complexity by focusing on crucial regions based on context. We leverage the target’s historical information stored in the spatial-temporal memory network to guide the tracker in better adapting to target deformation. Our model attains a 63.0% average overlap (AO) on the GOT-10K dataset, a 71.9% success score (AUC) on OTB2015, a 65.6% success score (AUC) on UAV123, and achieves 56.9% accuracy on the VOT2018 dataset. These outcomes demonstrate the high effectiveness of our proposed tracker in tracking a single object.


  • Robust Signboard Detection and Recognition in Real Environments

    by Pimpa Cheewaprakobkit, Chih-Yang Lin, Kuan-Hung Lin, Timothy K. Shih

    The detection and recognition of signboards have become increasingly important in the consumer electronics industry due to its wide range of potential applications. These applications include aiding visually impaired consumers in navigating through unfamiliar areas, identifying location landmarks for wayfinding, and providing targeted advertising and marketing services to consumers. However, the accuracy of signboard detection remains challenging due to the diversity of designs, which may incorporate text and images, and the complexity of environments, such as occlusion, shooting angles, and lighting conditions. Most existing detection methods struggle to distinguish small and similar signboards. In this paper, we propose robust signboard detection and recognition based on template generation. We also collected a new dataset that contains about 30,000 images, in 14 categories of signboards in Taiwan for training and free public use. The proposed method is a one-stage detector, which utilizes multi-scale features in the Darknet-19 network to learn object features effectively, detecting tiny and large objects. In addition, the proposed template generation method was designed to improve the overall accuracy. We compare our results with the Yolo series models. The results show that our proposed method more efficiently detects and recognizes signboards, achieving an mAP score of 95.99%, total parameters of 62.7M, and FPS of 8.3.