The growing attention towards immersive technologies such as augmented reality (AR), virtual reality (VR), mixed reality (MR), extended reality (XR), and the metaverse are revolutionizing cultural heritage education and tourism. Such technologies offer immersive and interactive experiences that transform the user’s exploration of museums, cultural heritage sites, educational content, and historical landmarks. This article presents a structured framework that addresses the advancement and application of these technologies in cultural heritage education to improve user experience, learning, emotional connection, and motivation. To further explore recent trends, issues, and opportunities, the article offers a comprehensive overview of the impact of state-of-the-art immersive technology on user experience within heritage education environments . The study also outlined standard questionnaires and effective methodologies for user experience evaluations. Furthermore, the article addresses the influence of standards and guidelines recommended by standardized bodies and organizations on XR and metaverse applications. It discussed how these standards and recommendations can play a role in setting protocols that shape the development of immersive heritage education environments. Finally, we introduce an architecture model for XR and metaverse applications that can assess developers, researchers, and stakeholders to enable immersive and interactive educational experiences, bridging geographical and physical barriers. This research is intended to help academic and industry stakeholders understand the integration of digital heritage preservation tools and user experience standards critical to advancing educational engagement in cultural heritage.

Pulse rate variability (PRV) analysis appears as the first alternative to heart rate variability analysis for wearable devices; however, there is a constraint on computational load and energy consumption for the limited system resources available to the devices. Considering that adjustment of the sampling frequency is one of the strategies for reducing computational load and power consumption, this study aimed to investigate the influence of sampling frequency (f s) on PRV analysis and to find the minimum sampling frequency while maintaining reliability. We generated 5000, 2500, 1000, 500, 250, 100, 50, 25, 20, 15, 10, 5 Hz down-sampled photoplethysmograms from 10 kHz-sampled PPGs and derived time- and frequency-domain variables of the PRV. These included AVNN, SDNN, SDSD, RMSSD, NN50, pNN50, total power, VLF, LF, HF, LF/HF, nLF and nHF for each down-sampled signal. Derived variables were compared with heart rate variability of the 10 kHz-sampled electrocardiograms, and then statistically investigated using one-way ANOVA test and Bland-Altman analysis. As a result, significant differences (P < 0.05) were found for SDNN, SDSD, RMSSD, NN50, pNN50, TP, HF, LF/HF, nLF and nHF, but not for AVNN, VLF and LF. Based on the post hoc tests, it was found that the NN50 and pNN50, SDSD and RMSSD, LF/HF and nHF, SDNN, TP and nLF analysis had significant differences at f s ⩽ 20 Hz, f s ⩽ 15 Hz, f s ⩽10 Hz; f s = 5 Hz, respectively. In other words, a significant difference was not seen for any variable if the f s was greater than 25 Hz. Consequently, our pilot study suggests that analysis of variability in the time and frequency domain from pulse rate obtained through PPG may be potentially as reliable as that derived from the analysis of the electrocardiogram, provided that f s ⩾25 Hz sampling frequency is used.

Recently, studies that analyze emotions based on physiological signals, such as electroencephalogram (EEG), by applying a deep learning algorithm have been actively conducted. However, the study of sequence modeling considering the change of emotional signals over time has not been fully investigated. To consider long-term interaction of emotion, in this study, we propose a long short-term memory network to consider changes in emotion over time and apply an attention mechanism to assign weights to the emotional states appearing at specific moments based on the peak–end rule in psychology. We used 32-channel EEG data from the DEAP database. Two-level (low and high) and three-level (low, middle, and high) classification experiments were performed on the valence and arousal emotion models. The results show accuracies of 90.1% and 87.9% using the two-level classification for the valence and arousal models with four-fold cross validation, respectively. In the case of the three-level classification, these values were obtained as 83.5% and 82.6%, respectively. Additional experiments were conducted using a network combining a convolutional neural network (CNN) submodule with the proposed model. The obtained results showed accuracies of 90.1% and 88.3% in the case of the two-level classification and 86.9% and 84.1% in the case of the three-level classification for the valence and arousal models with four-fold cross validation, respectively. In 10-fold cross validation, there were 91.8% for valence and 91.6% for arousal accuracy, respectively.

Detection of a brain tumor in the early stages is critical for clinical practice and survival rate. Brain tumors arise in multiple shapes, sizes, and features with various treatment options. Tumor detection manually is challenging, time-consuming, and prone to error. Magnetic resonance imaging (MRI) scans are mostly used for tumor detection due to their non-invasive properties and also avoid painful biopsy. MRI scanning of one patient’s brain generates many 3D images from multiple directions, making the manual detection of tumors very difficult, error-prone, and time-consuming. Therefore, there is a considerable need for autonomous diagnostics tools to detect brain tumors accurately. In this research, we have presented a novel TumorResnet deep learning (DL) model for brain detection, i.e., binary classification. The TumorResNet model employs 20 convolution layers with a leaky ReLU (LReLU) activation function for feature map activation to compute the most distinctive deep features. Finally, three fully connected classification layers are used to classify brain tumors MRI into normal and tumorous. The performance of the proposed TumorResNet architecture is evaluated on a standard Kaggle brain tumor MRI dataset for brain tumor detection (BTD), which contains brain tumor and normal MR images. The proposed model achieved a good accuracy of 99.33% for BTD. These experimental results, including the cross-dataset setting, validate the superiority of the TumorResNet model over the contemporary frameworks. This study offers an automated BTD method that aids in the early diagnosis of brain cancers. This procedure has a substantial impact on improving treatment options and patient survival.

Computer vision (CV) and human–computer interaction (HCI) are essential in many technological fields. Researchers in CV are particularly interested in real-time object detection techniques, which have a wide range of applications, including inspection systems. In this study, we design and implement real-time object detection and recognition systems using the single-shoot detector (SSD) algorithm and deep learning techniques with pre-trained models. The system can detect static and moving objects in real-time and recognize the object’s class. The primary goals of this research were to investigate and develop a real-time object detection system that employs deep learning and neural systems for real-time object detection and recognition. In addition, we evaluated the free available, pre-trained models with the SSD algorithm on various types of datasets to determine which models have high accuracy and speed when detecting an object. Moreover, the system is required to be operational on reasonable equipment. We tried and evaluated several deep learning structures and techniques during the coding procedure and developed and proposed a highly accurate and efficient object detection system. This system utilizes freely available datasets such as MS Common Objects in Context (COCO), PASCAL VOC, and Kitti. We evaluated our system’s accuracy using various metrics such as precision and recall. The proposed system achieved a high accuracy of 97% while detecting and recognizing real-time objects.

Deep learning (DL) has revolutionized image classification, yet deploying convolutional neural networks (CNNs) on edge devices for real-time applications remains a significant challenge due to constraints in computation, memory, and power efficiency. This work presents an optimized implementation of VGG16 and VGG19, two widely used CNN architectures, for classifying the CIFAR-10 dataset using transfer learning on field-programmable gate arrays (FPGAs). Utilizing the Xilinx Vitis-AI and TensorFlow2 frameworks, we adapt VGG16 and VGG19 for FPGA deployment through quantization, compression, and hardware-specific optimizations. Our implementation achieves high classification accuracy, with Top-1 accuracy of 89.54% and 87.47% for VGG16 and VGG19, respectively, while delivering significant reductions in inference latency (7.29× and 6.6× compared to CPU-based alternatives). These results highlight the suitability of our approach for resource-efficient, real-time edge applications. Key contributions include a detailed methodology for combining transfer learning with FPGA acceleration, an analysis of hardware resource utilization, and performance benchmarks. This work underscores the potential of FPGA-based solutions to enable scalable, low-latency DL deployments in domains such as autonomous systems, IoT, and mobile devices.