Automated recognition of human activities or actions has great significance as it incorporates wide-ranging applications, including surveillance, robotics, and personal health monitoring. Over the past few years, many computer vision-based methods have been developed for recognizing human actions from RGB and depth camera videos. These methods include space–time trajectory, motion encoding, key poses extraction, space–time occupancy patterns, depth motion maps, and skeleton joints. However, these camera-based approaches are affected by background clutter and illumination changes and applicable to a limited field of view only. Wearable inertial sensors provide a viable solution to these challenges but are subject to several limitations such as location and orientation sensitivity. Due to the complementary trait of the data obtained from the camera and inertial sensors, the utilization of multiple sensing modalities for accurate recognition of human actions is gradually increasing. This paper presents a viable multimodal feature-level fusion approach for robust human action recognition, which utilizes data from multiple sensors, including RGB camera, depth sensor, and wearable inertial sensors. We extracted the computationally efficient features from the data obtained from RGB-D video camera and inertial body sensors. These features include densely extracted histogram of oriented gradient (HOG) features from RGB/depth videos and statistical signal attributes from wearable sensors data. The proposed human action recognition (HAR) framework is tested on a publicly available multimodal human action dataset UTD-MHAD consisting of 27 different human actions. K-nearest neighbor and support vector machine classifiers are used for training and testing the proposed fusion model for HAR. The experimental results indicate that the proposed scheme achieves better recognition results as compared to the state of the art. The feature-level fusion of RGB and inertial sensors provides the overall best performance for the proposed system, with an accuracy rate of 97.6%.

Visually impaired persons (VIPs) comprise a significant portion of the population, and they are present around the globe and in every part of the world. In recent times, technology proved its presence in every domain, and innovative devices assist humans in their daily lives. In this work, a smart and intelligent system is designed for VIPs to assist mobility and ensure their safety. The proposed system provides navigation in real-time using an automated voice. Though VIPs wouldn’t be able to see objects in their surroundings, they can sense and visualize the roaming environment. Moreover, a web-based application is developed to ensure their safety. The user of this application can turn the on-demand function for sharing his/her location with the family while compromising privacy. Through this application, the family members of VIPs would be able to track their movement (get location and snapshots) while being at their homes. Hence, the device allows VIPs to visualize the environment and ensure their security. Such a comprehensive device was a missing link in the existing literature. The application uses MobileNet architecture due to its low computational complexity to run on low-power end devices. To assess the efficacy of the proposed system, six pilot studies have been performed that reflected satisfactory results. For object detection and recognition, a deep Convolution Neural Network (CNN) model is employed with an accuracy of 83.3%, whereas the dataset contains more than 1000 categories. Moreover, a score-based quantitative comparative analysis is performed using the supported features of devices. It is found that the proposed system has outperformed the existing devices having a total score of 9.1/10, which is 8% higher than the second-best.

The use of aerial base stations, AI cloud, and satellite storage can help manage location, traffic, and specific application-based services for vehicular social networks. However, sharing of such data makes the vehicular network vulnerable to data and privacy leakage. In this regard, this article proposes an efficient and secure data sharing scheme using community segmentation and a blockchain-based framework for vehicular social networks. The proposed work considers similarity matrices that employ the dynamics of structural similarity, modularity matrix, and data compatibility. These similarity matrices are then passed through stacked autoencoders that are trained to extract encoded embedding. A density-based clustering approach is then employed to find the community segments from the information distances between the encoded embeddings. A blockchain network based on the Hyperledger Fabric platform is also adopted to ensure data sharing security. Extensive experiments have been carried out to evaluate the proposed data-sharing framework in terms of the sum of squared error, sharing degree, time cost, computational complexity, throughput, and CPU utilization for proving its efficacy and applicability. The results show that the CSB framework achieves a higher degree of SD, lower computational complexity, and higher throughput.

The past decade has seen a rapid adoption of Artificial Intelligence (AI), specifically the deep learning networks, in Internet of Medical Things (IoMT) ecosystem. However, it has been shown recently that the deep learning networks can be exploited by adversarial attacks that not only make IoMT vulnerable to the data theft but also to the manipulation of medical diagnosis. The existing studies consider adding noise to the raw IoMT data or model parameters which not only reduces the overall performance concerning medical inferences but also is ineffective to the likes of deep leakage from gradients method. In this work, we propose proximal gradient split learning (PSGL) method for defense against the model inversion attacks. The proposed method intentionally attacks the IoMT data when undergoing the deep neural network training process at client side. We propose the use of proximal gradient method to recover gradient maps and a decision-level fusion strategy to improve the recognition performance. Extensive analysis show that the PGSL not only provides effective defense mechanism against the model inversion attacks but also helps in improving the recognition performance on publicly available datasets. We report 14.0 $\%$ , 17.9 $\%$ , and 36.9 $\%$ gains in accuracy over reconstructed and adversarial attacked images, respectively.