More Like this

1. The Impact of Pause and Filler Word Encoding on Dementia Detection with Contrastive Learning

   https://doi.org/10.3390/app14198879  

   Soleimani, Reza; Guo, Shengjie; Haley, Katarina L; Jacks, Adam; Lobaton, Edgar (October 2024, Applied Sciences)

   Dementia is primarily caused by neurodegenerative diseases like Alzheimer’s disease (AD). It affects millions worldwide, making detection and monitoring crucial. This study focuses on the detection of dementia from speech transcripts of controls and dementia groups. We propose encoding in-text pauses and filler words (e.g., “uh” and “um”) in text-based language models and thoroughly evaluating their impact on performance (e.g., accuracy). Additionally, we suggest using contrastive learning to improve performance in a multi-task framework. Our results demonstrate the effectiveness of our approaches in enhancing the model’s performance, achieving 87% accuracy and an 86% f1-score. Compared to the state of the art, our approach has similar performance despite having significantly fewer parameters. This highlights the importance of pause and filler word encoding on the detection of dementia. 

   more » « less
2. Enhancing Wearable Fall Detection System via Synthetic Data

   https://doi.org/10.3390/s25154639  

   Debnath, Minakshi; Alamgeer, Sana; Kabir, Md Shahriar; Ngu, Anne H (August 2025, Sensors)

   Deep learning models rely heavily on extensive training data, but obtaining sufficient real-world data remains a major challenge in clinical fields. To address this, we explore methods for generating realistic synthetic multivariate fall data to supplement limited real-world samples collected from three fall-related datasets: SmartFallMM, UniMib, and K-Fall. We apply three conventional time-series augmentation techniques, a Diffusion-based generative AI method, and a novel approach that extracts fall segments from public video footage of older adults. A key innovation of our work is the exploration of two distinct approaches: video-based pose estimation to extract fall segments from public footage, and Diffusion models to generate synthetic fall signals. Both methods independently enable the creation of highly realistic and diverse synthetic data tailored to specific sensor placements. To our knowledge, these approaches and especially their application in fall detection represent rarely explored directions in this research area. To assess the quality of the synthetic data, we use quantitative metrics, including the Fréchet Inception Distance (FID), Discriminative Score, Predictive Score, Jensen–Shannon Divergence (JSD), and Kolmogorov–Smirnov (KS) test, and visually inspect temporal patterns for structural realism. We observe that Diffusion-based synthesis produces the most realistic and distributionally aligned fall data. To further evaluate the impact of synthetic data, we train a long short-term memory (LSTM) model offline and test it in real time using the SmartFall App. Incorporating Diffusion-based synthetic data improves the offline F1-score by 7–10% and boosts real-time fall detection performance by 24%, confirming its value in enhancing model robustness and applicability in real-world settings. 

   more » « less
3. ARE GENERATIVE ADVERSARIAL NETWORKS CAPABLE OF GENERATING NOVEL AND DIVERSE DESIGN CONCEPTS? AN EXPERIMENTAL ANALYSIS OF PERFORMANCE

   https://doi.org/10.1017/pds.2023.64  

   Ghasemi, Parisa; Yuan, Chenxi; Marion, Tucker; Moghaddam, Mohsen (July 2023, Proceedings of the Design Society)

   Generative Adversarial Networks (GANs) have shown stupendous power in generating realistic images to an extend that human eyes are not capable of recognizing them as synthesized. State-of-the-art GAN models are capable of generating realistic and high-quality images, which promise unprecedented opportunities for generating design concepts. Yet, the preliminary experiments reported in this paper shed light on a fundamental limitation of GANs for generative design: lack of novelty and diversity in generated samples. This article conducts a generative design study on a large-scale sneaker dataset based on StyleGAN, a state-of-the-art GAN architecture, to advance the understanding of the performance of these generative models in generating novel and diverse samples (i.e., sneaker images). The findings reveal that although StyleGAN can generate samples with quality and realism, the generated and style-mixed samples highly resemble the training dataset (i.e., existing sneakers). This article aims to provide future research directions and insights for the engineering design community to further realize the untapped potentials of GANs for generative design. 

   more » « less
4. Leveraging GANs for Synthetic Data Generation to Improve Intrusion Detection Systems

   https://doi.org/10.62411/faith.3048-3719-52  

   Rahman, Md Abdur; Francia, Guillermo A; Shahriar, Hossain (March 2025, Journal of Future Artificial Intelligence and Technologies)

   This research presents a hybrid intrusion detection approach that integrates Generative Adversarial Networks (GANs) for synthetic data generation with Random Forest (RF) as the primary classifier. The study aims to improve detection performance in cybersecurity applications by enhancing dataset diversity and addressing challenges in traditional models, particularly in detecting minority attack classes often underrepresented in real-world datasets. The proposed method employs GANs to generate synthetic attack samples that mimic real-world intrusions, which are then combined with real data from the UNSW-NB15 dataset to create a more balanced training set. By leveraging synthetic data augmentation, our approach mitigates issues related to class imbalance and enhances the generalization capability of the classifier. Extensive experiments demonstrate that RF trained on the combined dataset of real and synthetic data achieves superior detection performance compared to models trained exclusively on real data. Specifically, RF trained solely on the original dataset achieves an accuracy of 97.58%, whereas integrating GAN-generated synthetic data improves accuracy to 98.27%. The proposed methodology is further evaluated through comparative analysis against alternative classifiers, including Support Vector Machine (SVM), XGBoost, Gated Recurrent Unit (GRU), and related studies in the field. Our findings indicate that GAN-augmented training significantly enhances detection rates, particularly for rare attack types, while maintaining computational efficiency. Furthermore, RF outperforms other classifiers, including deep learning models, demonstrating its effectiveness as a lightweight yet robust classification method. Integrating GANs with RF offers a scalable and adaptable framework for intrusion detection, ensuring improved resilience against evolving cyber threats. 

   more » « less
5. Advancing ai-powered medical image synthesis: Insights from medvqa-gi challenge using clip, fine-tuned stable diffusion, and dream-booth + lora

   Ojonugwa, E; Rahman, M; Khalifa, F (September 2024, CEUR workshop proceedings)

   Faggioli, G; Ferro, N; Galuščáková, P; Herrera, A (Ed.)

   The MEDVQA-GI challenge addresses the integration of AI-driven text-to-image generative models in medical diagnostics, aiming to enhance diagnostic capabilities through synthetic image generation. Existing methods primarily focus on static image analysis and lack the dynamic generation of medical imagery from textual descriptions. This study intends to partially close this gap by introducing a novel approach based on fine-tuned generative models to generate dynamic, scalable, and precise images from textual descriptions. Particularly, our system integrates fine-tuned Stable Diffusion and DreamBooth models, as well as Low-Rank Adaptation (LORA), to generate high-fidelity medical images. The problem is around two sub-tasks namely: image synthesis (IS) and optimal prompt production (OPG). The former creates medical images via verbal prompts, whereas the latter provides prompts that produce high-quality images in specified categories. The study emphasizes the limitations of traditional medical image generation methods, such as hand sketching, constrained datasets, static procedures, and generic models. Our evaluation measures showed that Stable Diffusion surpasses CLIP and DreamBooth + LORA in terms of producing high-quality, diversified images. Specifically, Stable Diffusion had the lowest Fréchet Inception Distance (FID) scores (0.099 for single center, 0.064 for multi-center, and 0.067 for combined), indicating higher image quality. Furthermore, it had the highest average Inception Score (2.327 across all datasets), indicating exceptional diversity and quality. This advances the field of AI-powered medical diagnosis. Future research will concentrate on model refining, dataset augmentation, and ethical considerations for efficiently implementing these advances into clinical practice. 

   more » « less