Volume 9, issue 6

As society ages, the imbalance between family caregivers and elderly individuals increases, leading to inadequate support for seniors in many regions. This situation has ignited interest in automatic health monitoring systems, particularly in fall detection, due to the significant health risks that falls pose to older adults. This research presents a vision-based fall detection system that employs computer vision and deep learning to improve elderly care. Traditional systems often struggle to accurately detect falls from various camera angles, as they typically rely on static assessments of body posture. To tackle this challenge, we implement a feature fusion strategy within a deep learning framework to enhance detection accuracy across diverse perspectives. The process begins by generating a Human Silhouette Image (HSI) through background subtraction. By combining silhouette images from two consecutive frames, we create a Silhouette History Image (SHI), which captures the shape features of the individual. Simultaneously, Dense Optical Flow (DOF) extracts motion features from the same frames, allowing us to merge these with the SHI for a comprehensive input image. This fused representation is then processed using a pre-trained Convolutional Neural Network (CNN) to extract deep features. A Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) is subsequently trained on these features to recognize patterns indicative of fall events. Our approach’s effectiveness is validated through experiments on the UP-fall detection dataset, which includes 1,122 action videos and achieves an impressive 99% accuracy in fall detection.

As society ages, the imbalance between family caregivers and elderly individuals increases, leading to inadequate support for seniors in many regions. This situation has ignited interest in automatic health monitoring systems, particularly in fall detection, due to the significant health risks that falls pose to older adults. This research presents a vision-based fall detection system that employs computer vision and deep learning to improve elderly care. Traditional systems often struggle to accurately detect falls from various camera angles, as they typically rely on static assessments of body posture. To tackle this challenge, we implement a feature fusion strategy within a deep learning framework to enhance detection accuracy across diverse perspectives. The process begins by generating a Human Silhouette Image (HSI) through background subtraction. By combining silhouette images from two consecutive frames, we create a Silhouette History Image (SHI), which captures the shape features of the individual. Simultaneously, Dense Optical Flow (DOF) extracts motion features from the same frames, allowing us to merge these with the SHI for a comprehensive input image. This fused representation is then processed using a pre-trained Convolutional Neural Network (CNN) to extract deep features. A Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) is subsequently trained on these features to recognize patterns indicative of fall events. Our approach’s effectiveness is validated through experiments on the UP-fall detection dataset, which includes 1,122 action videos and achieves an impressive 99% accuracy in fall detection.

This study proposes the Value Karuta (VK), an application of the traditional Japanese card game karuta. Its goal is to contribute to the understanding of value, which is the first principle of lean management. After stating the problems of lean management and the specifications of VK, this paper confirms the validity of the proposal by discussing two surveys. The first survey explored the utility factors of the cards themselves; it was conducted with a group of students and businesspeople in Japan. The second survey observed the actual situation in the game and was conducted in a group of academics at two UK universities. Both surveys used qualitative methods, such as observation and discussion, and quantitative questionnaires. The results confirm the role of VK as a fundamental tool addressing the need to understand customer value in lean management.

Deep Neural Networks (DNNs) prove to be susceptible to synthetically generated samples, so-called adversarial examples. Such adversarial examples aim at generating misclassifications by specifically optimizing input data for a matching perturbation. With the increasing use of deep learning on embedded devices and the resulting use of quantization techniques to compress deep neural networks, it is critical to investigate the adversarial vulnerability of quantized neural networks.In this paper, we perform an in-depth study of the adversarial robustness of quantized networks against direct attacks, where adversarial examples are both generated and applied on the same network. Our experiments show that quantization makes models resilient to the generation of adversarial examples, even for attacks that demonstrate a high success rate, indicating that it offers some degree of robustness against these attacks. Additionally, we open-source Adversarial Neural Network Toolkit (ANNT) to support the replication of our results.