Journal of Ambient Intelligence and Humanized Computing最新文献

Security and privacy are fundamental to applications of medical internet of things (IoT). This article proposes a new computed tomography (CT) image three-classification prediction network, Re50-ViT (ResNet50 and Vision Transformer), which aims to improve the accuracy of traditional neural networks in screening patients with novel coronavirus infection pneumonia. To enhance network performance, the batch normalization layer is replaced with the group normalization layer for more stable activation normalization. The front-end utilizes ResNet50 for local feature extraction, and global information integration is achieved through the connection of a Class token and position embedding. Dropout layer is added to prevent overfitting and improve generalization. multiple transformer encoder layers are used to capture complex patterns and model label relationships within the CT images. The network integrates human-centric IoT and security measures to protect patient privacy and sensitive medical information. Experimental results compared to existing methods demonstrate the superiority of the Re50-ViT network. The Grad-CAM (gradient-weighted class activation mapping) technique provides intuitive visualization, highlighting the importance of specific regions in the CT images. The network shows effectiveness and reliability in detecting lung lesions, including COVID-19 and other pulmonary abnormalities. The integration of human-centric IoT and security considerations further enhances the clinical value of the network while ensuring the protection of patient data and privacy.

Predicting merger waves has been a classical yet challenging problem. In this paper, we propose approaches to predict industry merger waves relying on an integrated dataset including financial statements and supply data, as well as more than 60 thousand firm-level mergers and acquisitions records. We utilize 1000-dimension features—including common-used industry characteristics and novel supply network information—for predictions and train classifiers based on different machine learning methods. The experiments demonstrate the usefulness of our prediction approach, as the predicting precision reaches 91% on acquirers and 96% on targets. By further analysis, some patterns are well explained by financial theories, such as the well-known Tobin’s Q measurement. Especially, new influential factors on merger waves are revealed by the empirical analysis on micro-structure network features. To the best of our knowledge, this paper is one of the first attempts to explore merger waves prediction, and our approaches and findings introduce a new viewpoint for this field.

In order to solve the problem of low detection efficiency and long working time in the traditional video surveillance system for abnormal behavior detection and identification methods. A multimodal abnormal behavior detection and identification method based on video surveillance is proposed and applied to an online video classroom concentration evaluation task for college students in English. The model works by capturing abnormal behaviors and facial expressions and building a joint network that fuses abnormal behaviors and facial expressions. By testing on two open-source datasets and self-built classroom real-time datasets, the results verify that the model in this paper has better recognition performance compared to current mainstream models while maintaining real-time performance. The model proposed in this paper provides a new way of thinking about building smart classrooms.

This article proposes a new intelligent trajectory tracking control law for the precise maneuvering of an autonomous vehicle in the presence of parametric uncertainties and external disturbances. The controller design includes a fuzzy sliding mode algorithm for smooth motion control subjected to steering saturation and curvature constraints. Along with the Salp Swarm Optimization technique, explored for optimal selection of surface coefficient in fractional order Proportional-Derivative type (P{D}^{alpha }) sliding manifold. The sliding variable on the surface approaches zero in a finite time. Further, the trajectory tracking control rule offers the stability of closed-loop tracking on the predetermined path and ensures finite time convergence to the sliding surface. In addition, to estimate the hitting gain in online mode, a supervisory fuzzy logic controller system is used. Therefore, it is not necessary to determine upper bounds on uncertainty in the dynamic parameters of autonomous vehicles. Lyapunov theory verifies the global asymptotic stability of the entire closed-loop control strategy. The major control issue is the input constraints arising primarily due to the capability of the steering actuating module, which causes significant deviation or vehicle instability. Consequently, it is desirable to design a robust adaptive stable controller, such as Adaptive Backstepping Control (ABC), even though it requires vehicle model information. Therefore, the proposed model-free intelligent sliding mode technique offers better tracking performance and vehicle stability in adverse conditions. Finally, the efficacy of the proposed control technique was confirmed through a comparative analysis based on numerical simulation using MATLAB/SIMULINK and experimental validation using Quanser’s self-driving car module. A quantitative study was conducted to elucidate the superior tracking performance of intelligent control over the traditional SMC and adaptive backstepping control methods.

Date fruit, a vital agricultural product in the Middle East area, is harvested annually in millions of metric tons and is renowned for its abundant nutrients. With computer vision and machine learning techniques, automatic date fruit classification enables farmers and supermarkets to differentiate between various varieties and qualities of date fruits within their inventory. Date fruits have unique physical characteristics, such as shape, size, color, texture, and skin type that are important in determining their variety and quality. These characteristics can vary significantly depending on the cultivar, growing conditions, and ripening stage of the date fruits. This paper presents a novel date fruit type classification and grading system achieved through the feature-level fusion of deep learning features and wavelet scattering features. Wavelet scattering features are extracted at varying levels of decomposition; enabling reliable extraction of information from diverse channels. To extract deep features this study utilizes pre-trained architectures, including Alexnet, Googlenet, Resnet, and MobileNetV2. The proposed methodology has been experimentally evaluated with the Date Fruit in Controlled Environment dataset, which has nine classes, and has yielded an accuracy of 95.9% for date species classification. Various date fruit species from the TU-DG dataset were graded, and for Ajwa species, the accuracy is 97.8%, for Mabroom, 92.6% accuracy, and for Sukkary, 99.5% accuracy.

Gait recognition can exploit the signals from wearables, e.g., the accelerometers embedded in smart devices. At present, this kind of recognition mostly underlies subject verification: the incoming probe is compared only with the templates in the system gallery that belong to the claimed identity. For instance, several proposals tackle the continuous recognition of the device owner to detect possible theft or loss. In this case, assuming a short time between the gallery template acquisition and the probe is reasonable. This work rather investigates the viability of a wider range of applications including identification (comparison with a whole system gallery) in the medium-long term. The first contribution is a procedure for extraction and two-phase selection of the most relevant aggregate features from a gait signal. A model is trained for each identity using Logistic Regression. The second contribution is the experiments investigating the effect of the variability of the gait pattern in time. In particular, the recognition performance is influenced by the benchmark partition into training and testing sets when more acquisition sessions are available, like in the exploited ZJU-gaitacc dataset. When close-in-time acquisition data is only available, the results seem to suggest re-identification (short time among captures) as the most promising application for this kind of recognition. The exclusive use of different dataset sessions for training and testing can rather better highlight the dramatic effect of trait variability on the measured performance. This suggests acquiring enrollment data in more sessions when the intended use is in medium-long term applications of smart ambient intelligence.

The expected spatial coverage of a crowdsensing platform is an important parameter that derives from the mobility data of the crowdsensing platform users. We tackle the challenge of estimating the anticipated coverage while adhering to privacy constraints, where the platform is restricted from accessing detailed mobility data of individual users. Specifically, we model the coverage as the probability that a user detours to a point of interest if the user is present in a certain region around that point. Following this approach, we propose and evaluate a centralized as well as a distributed implementation model. We examine real-world mobility data employed for assessing the coverage performance of the two models, and we show that the two implementation models provide different privacy requirements but are equivalent in terms of their outputs.

Currently, the move from traditional healthcare to smart healthcare systems is greatly aided by current technology. Healthcare proposes a new healthcare model that is patient-centered using advancements in wearable sensors, connectivity, and the Internet of Things (IoT). The administration of enormous amounts of data, including reports and pictures of every individual, increases human labour requirements and security hazards. This study shows how a blockchain-based Internet of Things might improve patient care while lowering costs by using medical resources more wisely. Initially, Resource Provider’s IoT data will be sensed and encrypts using Diffie Hellman Galois–Elliptic-curve cryptography (DHG-ECC). Next, from the extracted attributes, the optimal features will be selected by using Pearson Correlation Coefficient based Sand Cat Optimization Algorithm (PCC-SCOA). After that, the selected optimal features will be combined and converted into hashcode using the Digit Folding–Streebog Hashing algorithm. This hashcode will be constructed in the form of Smart Contract. Next, the Resource Requester (Doctor or Nurse) sends the Role Request with the Combined Linear Congruential Generator–Digital Signature Algorithm (CLCG-DSA). The next Resource Requester will be matching the hashed access policy with Blockchain. The proposed models are used to compare the performance of proposed design using feature selection time, Encryption time, Decryption time, security level, signature creation time and signature verification time. Our proposed method DHGECC approach achieves 96.123% higher security.

A W-operator is an image transformation that is locally defined inside a window W, invariant to translations. The automatic design of the W-operators consists of the design of functions, whose domain is a set of patterns or vectors obtained by translating a window through training images and the output of each vector is a class or label. The main difficulty to consider when designing W-operators is the generalization problem that occurs due to lack of training images. In this work, we propose the use of membership functions to solve the generalization problem in gray level images. Membership functions are defined from the training images to model regions that are often inaccurate due to ambiguous gray levels in the images. This proposal was applied to brain magnetic resonance image segmentation to test its performance in a field of interest in biomedical images. The experiments were carried out with different numbers of training and test images, windows sizes of (3times 3), (5times 5), (7times 7), (11times 11), and (15times 15), and images with noise levels at 0, 1, 3, 5, 7, and 9(%). To calculate the performance of each designed W-operator, the classification error, sensitivity, and specificity were used. From the experimental results, it was concluded that the best performance is achieved with a window of size (3times 3). In images with noise levels from 1 to 5(%), the classification error is less than 4(%) and the sensitivity and specificity are greater than 94 and 98(%), respectively.