Search for: All records

Accurate weight measurement is critical in emergency medicine, particularly for the precise administration of medications and treatments. However, traditional methods of weight estimation can be unreliable, especially in time-sensitive or resource-limited environments. This study provides a comprehensive review of the advancements and techniques in body weight estimation, with a focus on modern approaches leveraging contactless sensors, such as 3D cameras, and AI-powered computational models. The research evaluates the accuracy, reliability, and practical applicability of these methods across different contexts, including healthcare, forensic sciences, and emergency response. Additionally, this study identifies the limitations of current methodologies and uncovers gaps in the literature that warrant further investigation. Our findings aim to guide future research efforts and the development of more precise and scalable weight estimation solutions, ultimately enhancing their applicability in a variety of sectors. 

With the ongoing expansion of the aging population, it is increasingly critical to prioritize the safety of older drivers. The objective of this study is to utilize sensor data in order to detect early indications of impairment, thereby facilitating proactive interventions and enhancing road safety for the elderly. This article provides an overview of the research approach, presents significant results, and analyzes the consequences of utilizing in-vehicle sensors i.e. vision and telematics, to mitigate cognitive decline among elderly drivers; in doing so, it promotes progress in the domains of public health and transportation safety by standardizing the use of such devices to automatically assess the drivers’ cognitive functions. 

Acquiring labeled datasets often incurs substantial costs primarily due to the requirement of expert human intervention to produce accurate and reliable class labels. In the modern data landscape, an overwhelming proportion of newly generated data is unlabeled. This paradigm is especially evident in domains such as fraud detection and datasets for credit card fraud detection. These types of data have their own difficulties associated with being highly class imbalanced, which poses its own challenges to machine learning and classification. Our research addresses these challenges by extensively evaluating a novel methodology for synthesizing class labels for highly imbalanced credit card fraud data. The methodology uses an autoencoder as its underlying learner to effectively learn from dataset features to produce an error metric for use in creating new binary class labels. The methodology aims to automatically produce new labels with minimal expert input. These class labels are then used to train supervised classifiers for fraud detection. Our empirical results show that the synthesized labels are of high enough quality to produce classifiers that significantly outperform a baseline learner comparison when using area under the precision-recall curve (AUPRC). We also present results of varying levels of positive-labeled instances and their effect on classifier performance. Results show that AUPRC performance improves as more instances are labeled positive and belong to the minority class. Our methodology thereby effectively addresses the concerns of high class imbalance in machine learning by creating new and effective class labels.