The definition of human-computer interaction (HCI) has changed in the current year because people are interested in their various ergonomic devices ways. Many researchers have been working to develop a hand gesture recognition system with a kinetic sensor-based dataset, but their performance accuracy is not satisfactory. In our work, we proposed a multistage spatial attention-based neural network for hand gesture recognition to overcome the challenges. We included three stages in the proposed model where each stage is inherited the CNN; where we first apply a feature extractor and a spatial attention module by using self-attention from the original dataset and then multiply the feature vector with the attention map to highlight effective features of the dataset. Then, we explored features concatenated with the original dataset for obtaining modality feature embedding. In the same way, we generated a feature vector and attention map in the second stage with the feature extraction architecture and self-attention technique. After multiplying the attention map and features, we produced the final feature, which feeds into the third stage, a classification module to predict the label of the correspondent hand gesture. Our model achieved 99.67%, 99.75%, and 99.46% accuracy for the senz3D, Kinematic, and NTU datasets.

Fitness is important in people’s lives. Good fitness habits can improve cardiopulmonary capacity, increase concentration, prevent obesity, and effectively reduce the risk of death. Home fitness does not require large equipment but uses dumbbells, yoga mats, and horizontal bars to complete fitness exercises and can effectively avoid contact with people, so it is deeply loved by people. People who work out at home use social media to obtain fitness knowledge, but learning ability is limited. Incomplete fitness is likely to lead to injury, and a cheap, timely, and accurate fitness detection system can reduce the risk of fitness injuries and can effectively improve people’s fitness awareness. In the past, many studies have engaged in the detection of fitness movements, among which the detection of fitness movements based on wearable devices, body nodes, and image deep learning has achieved better performance. However, a wearable device cannot detect a variety of fitness movements, may hinder the exercise of the fitness user, and has a high cost. Both body-node-based and image-deep-learning-based methods have lower costs, but each has some drawbacks. Therefore, this paper used a method based on deep transfer learning to establish a fitness database. After that, a deep neural network was trained to detect the type and completeness of fitness movements. We used Yolov4 and Mediapipe to instantly detect fitness movements and stored the 1D fitness signal of movement to build a database. Finally, MLP was used to classify the 1D signal waveform of fitness. In the performance of the classification of fitness movement types, the mAP was 99.71%, accuracy was 98.56%, precision was 97.9%, recall was 98.56%, and the F1-score was 98.23%, which is quite a high performance. In the performance of fitness movement completeness classification, accuracy was 92.84%, precision was 92.85, recall was 92.84%, and the F1-score was 92.83%. The average FPS in detection was 17.5. Experimental results show that our method achieves higher accuracy compared to other methods.

3D-Network-on-Chips exploit the benefits of Network-on-Chips and 3D-Integrated Circuits allowing them to be considered as one of the most advanced and auspicious communication methodologies. On the other hand, the reliability of 3D-NoCs, due to the vulnerability of Through Silicon Vias, remains a major problem. Most of the existing techniques rely on correcting the TSV defects by using redundancies or employing routing algorithms. Nevertheless, they are not suitable for TSV-cluster defects as they can either lead to costly area and power consumption overheads, or they may result in non-minimal routing paths; thus, posing serious threats to the system reliability and overall performance. In this work, we present a scalable and low-overhead TSV usage and design method for 3D-NoC systems where the TSVs of a router can be utilized by its neighbors to deal with the cluster open defects. An adaptive online algorithm is also introduced to assist the proposed system to immediately work around the newly detected defects without using redundancies. The experimental results show the proposal ensure less than 2 percent of the routers being disabled, even with 50 percent of the TSV clusters defects. The performance evaluations also demonstrate unchanged performances for real applications under 5 percent of cluster defects.

The ample amount of information from hyperspectral image (HSI) bands allows the non-destructive detection and recognition of earth objects. However, dimensionality reduction (DR) of hyperspectral images (HSI) is required before classification as the classifier may suffer from the curse of dimensionality. Therefore, dimensionality reduction plays a significant role in HSI data analysis (e.g., effective processing and seamless interpretation). In this article, a sophisticated technique established as t-Distributed Stochastic Neighbor Embedding (tSNE) following the dimension reduction along with a blended CNN was implemented to improve the visualization and characterization of HSI. In the procedure, first, we employed principal component analysis (PCA) to reduce the HSI dimensions and remove non-linear consistency features between the wavelengths to project them to a smaller scale. Then we proposed tSNE to preserve the local and global pixel relationships and check the HSI information visually and experimentally. Lastly, it yielded two-dimensional data, improving the visualization and classification accuracy compared to other standard dimensionality-reduction algorithms. Finally, we employed deep-learning-based CNN to classify the reduced and improved HSI intra- and inter-band relationship-feature vector. The evaluation performance of $95.21\%$ accuracy and $6.2\%$ test loss proved the superiority of the proposed model compared to other state-of-the-art DR reduction algorithms.

Neurological disorders, including Brain Tumors (BTs), Alzheimer's Disease (AD), and Parkinson's Disease (PD), pose significant global health challenges. Early and accurate diagnosis is crucial for effective treatment and improved patient outcomes. Magnetic Resonance Imaging (MRI) is a key diagnostic tool, but traditional Machine Learning (ML) approaches often rely on labor-intensive handcrafted features, leading to inconsistent performance. Recent advancements in Deep Learning (DL) enable automated feature extraction, which offers improved robustness and scalability. However, many existing methods face challenges in fully exploiting the complementary strengths of DL and handcrafted features across multiple disease types. This study proposes a novel hybrid DL model that integrates automated deep features with statistical textural descriptors for the classification of BTs, AD, and PD. The model employs a dual-stream architecture: (1) a modified VGG16 convolutional neural network (CNN), chosen for its favorable trade-off between performance and computational efficiency in medical imaging, to extract deep features from MRI slices, and (2) a sequential 1D CNN to process six GLCM-derived handcrafted features, empirically validated for their superior discriminative power in neuroanatomical texture analysis. By integrating these complementary feature sets, the model leverages global patterns and fine-grained textural details, resulting in a robust and comprehensive representation for accurate and reliable medical image classification. Grad-CAM is incorporated to enhance interpretability by localizing diagnostically relevant brain regions. The fused features are passed through a fully connected layer for final classification. The proposed model was evaluated on four publicly available MRI datasets, achieving accuracies of 98.86%, 99.50%, 98.52%, and 99.13% on the CE-MRI (multi-class BT), Br35H (binary BT), AD, and PD datasets, respectively. The model achieved an average classification accuracy of 99.05% across the three neurological disorders. Our method outperforms recent state-of-the-art (SOTA) methods, which shows the effectiveness of the proposed model integrating DL and handcrafted features to develop interpretable, robust, and generalizable AI-driven diagnostic systems.