Autonomous navigation in crowded environments remains a significant challenge due to the highly dynamic and unpredictable nature of pedestrian movements. This paper presents a novel approach for socially-compliant crowd navigation by leveraging human pose tracking, trajectory prediction, and obstacle avoidance techniques. We introduce PoseTrajNet, an end-to-end autonomous agent navigation pipeline that integrates YOLOv8 for object detection, BlazePose for real-time human pose estimation, and a custom trajectory prediction model drawing on concepts from Social GANs. PoseTrajNet employs pose keypoints as socially-compliant features to anticipate pedestrian trajectories, enabling proactive path planning and dynamic safe radius adjustments for obstacle avoidance. Extensive evaluations on standard datasets demonstrate PoseTrajNet's effectiveness in seamless crowd navigation, outperforming baselines while adhering to social norms.

In contemporary diagnostic workflows, medical image captioning has emerged as a pivotal advancement, combining deep learning methodologies and transformer architectures to enhance accuracy and efficiency in medical interpretations. This paper proposes the optimization of multimodal transformers for automated medical image captioning, focusing on integrating Vision Transformers (ViT) and Bidirectional Auto-Regressive Transformers (BART) with novel variations such as Swin Transformers and GPT-2. We use a robust multimodal AI framework to explore how these architectures synergize to generate coherent and diagnostically relevant captions for radiological images. We assess the performance of multiple transformer models by employing the ROCO dataset, containing paired X-ray images and expert-generated reports. Our findings demonstrate that the ViT + BART combination yields the most stable and accurate captions, minimizing training and validation loss. In contrast, DEiT + MBART displayed instability, highlighting the need for further hyperparameter tuning. Through this comparative analysis, we underscore the critical role of transformer-based models in reducing the cognitive load on medical professionals, enhancing diagnostic accuracy, and promoting real-time, automated medical image interpretation.

Overfitting remains a persistent challenge in deep learning. It is primarily attributed to data outliers, noise, and limited training set sizes. This paper presents Divide2Conquer (D2C), a novel technique designed to address this issue. D2C proposes partitioning the training data into multiple subsets and training separate identical models on them. To avoid overfitting on any specific subset, the trained parameters from these models are aggregated and averaged periodically throughout the training phase, enabling the model to learn from the entire dataset while mitigating the impact of individual outliers or noise. Empirical evaluations on multiple benchmark datasets across various deep learning tasks demonstrate that D2C effectively improves generalization performance, particularly for larger datasets. This study verifies D2C’s ability to achieve significant performance gains both as a standalone technique and when used in conjunction with other overfitting reduction methods through a series of experiments, including analysis of decision boundaries, loss curves, and other performance metrics. It also provides valuable insights into the implementation and hyperparameter tuning of D2C. Our codes are publicly available at: https://github.com/Saiful185/Divide2Conquer.

The widespread use of social media platforms has led to an increase in the dissemination of fake news with the intention of manipulating public opinion and causing chaos and panic among the population. To address this issue, we focus on detecting the organized groups that participate together in fake news campaigns without prior knowledge of the news content or the profiles of social accounts. To this end, we propose aspatial–temporal similarity graph, a novel graph structure that connects social accounts that participate in the early stage of similar fake news campaigns. A community detection algorithm is applied on the similarity graph to cluster the users into communities. We propose acommunity labeling algorithmto label the communities as benign or malicious based on the output of a fake news classifier. Evaluation results show that the community labeling algorithm can correctly label the communities with an accuracy of$99.61\%$. In addition, we perform a statistical comparison analysis to identify the structural community features that are statistically significant between benign and malicious communities.

This article introduces a novelattentive drivingframework in intelligent transportation systems (ITS) to investigate the influence of cognitive behavior on distracting driving activities that lead to inattention while driving. Therefore, this work proposes a holistic computational and communication framework that can monitor on-compartment real-time multimodal sensory observation such as physiological, camera, and environmental inputs while capable of distraction detection and emotion recognition for driver's mood stabilization. In particular, this work develops a capsule network for distraction detection, a 1-D convolutional neural network for emotion recognition, an a priori algorithm for sequential context fusion, and a Bayesian network for recommending auditory stimulus content for driver mood stabilization and audio-visual safety messages for road safety. Further, an asynchronous client control scheme has developed to overcome the challenges of multitime scale sensory observations and communicate among the multimodel sensory hubs. Finally, a prototype is developed and tested in a simulation environment. The quantitative analysis results show that the proposed framework can successfully detect around 89% and 87% of distractive activities and the affective state of a driver, respectively. Finally, based on experimental results, the proposed system demonstrates the capability to sustain a driver's attention for approximately 97% of the time, with a confidence level of 95%.

Abstract Human conversational styles are measured by the sense of humor, personality, and tone of voice. These characteristics have become essential for conversational intelligent virtual assistants. However, most of the state-of-the-art intelligent virtual assistants (IVAs) are failed to interpret the affective semantics of human voices. This research proposes an anthropomorphic intelligent system that can hold a proper human-like conversation with emotion and personality. A voice style transfer method is also proposed to map the attributes of a specific emotion. Initially, the frequency domain data (Mel-Spectrogram) is created by converting the temporal audio wave data, which comprises discrete patterns for audio features such as notes, pitch, rhythm, and melody. A collateral CNN-Transformer-Encoder is used to predict seven different affective states from voice. The voice is also fed parallelly to the deep-speech, an RNN model that generates the text transcription from the spectrogram. Then the transcripted text is transferred to the multi-domain conversation agent using blended skill talk, transformer-based retrieve-and-generate generation strategy, and beam-search decoding, and an appropriate textual response is generated. The system learns an invertible mapping of data to a latent space that can be manipulated and generates a Mel-spectrogram frame based on previous Mel-spectrogram frames to voice synthesize and style transfer. Finally, the waveform is generated using WaveGlow from the spectrogram. The outcomes of the studies we conducted on individual models were auspicious. Furthermore, users who interacted with the system provided positive feedback, demonstrating the system’s effectiveness.

The traditional dietary recommendation systems are basically nutrition or health-aware where the human feelings on food are ignored. Human affects vary when it comes to food cravings, and not all foods are appealing in all moods. It takes a lot of effort to learn people’s food preferences and make recommendations based on their affects and nutrition. A questionnaire-based and preference-aware meal recommendation system can be a solution. However, automated recognition of social affects on different foods and planning the menu considering nutritional demand and social affect has some significant benefits over the questionnaire-based and preference-aware meal recommendations. A patient with severe illness, a person in a coma, or patients with locked-in syndrome and amyotrophic lateral sclerosis (ALS) cannot express their meal preferences. Therefore, the proposed framework includes a social-affective computing module to recognize the affects of different meals where the person’s affect is detected using electroencephalography (EEG) signals. EEG allows to capture the brain signals and analyze them to anticipate affective state toward a food. In this study, we have used a 14-channel wireless Emotiv Epoc+ to measure affectivity for different food items. A hierarchical ensemble method is applied to predict affectivity upon multiple feature extraction methods and Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) is used to generate a food list based on the predicted affectivity. In addition to the meal recommendation, an automated menu planning approach is also proposed considering a person’s energy intake requirement, affectivity, and nutritional values of the different menus. The bin-packing algorithm is used for the personalized menu planning of breakfast, lunch, dinner, and snacks. The experimental findings reveal that the suggested affective computing, meal recommendation, and menu planning algorithms perform well across a variety of assessment parameters.

The smartphone-based sensors (including accelerometer, proximity, and gyroscope sensors) are ubiquitous and emerging mobility data sources that could be used for transportation modes (i.e. bus, train, car, walking, and stationary) detection. One of the important challenges in transportation modes detection is to build an appropriate model that can extract useful data from the sensor outputs and that can reduce misclassifications. Several factors make the feature modeling difficult including inappropriate sampling frequency of input signals, wavering behavior of devices (e.g. the changing orientation of a device relative to the human body), and continuous base vibration causing similar sensor outputs for both stationary and non-stationary states and related threshold values of velocity. This paper proposes novel approaches to address these challenges by developing a robust transportation mode detector based on a convolution neural network (CNN). The proposed robust detector develops a feature modeling technique by novel feature fusion and cloning techniques. Pre-trained features are constructed using a separate vanilla neural network (VNN) framework to extract the distinguishing components from the original features that are combined with the original and cloned features. The proposed feature fusion technique is successfully able to overcome the noise from the base vibration and the minimal informative outputs from the lower sampling frequency. This enables the CNN to be trained with more efficient and discriminative features that result in a better classification model. The proposed approaches have been validated using a large volume of mobile sensor data based on the movements of travelers. Different types of mobile sensors have been used to collect data including accelerometer, proximity, and gyroscope. Experimental results demonstrate that the proposed approaches can improve the performance of the detection engine significantly over conventional techniques and reduces the misclassification rate.

Cryptocurrencies are the new form of trade that has revolutionized how we look into our financial institutions. Bitcoin dominates the industry with the highest market share among the hundreds of other cryptocurrencies. However, high energy consumption leading to increasing carbon emission, prioritizing high-value transactions, and long waiting times are some of the flaws preventing it from reaching its full potential. Owing to the block rewards getting halved every four years, miners and researchers are fearful that this would be the breaking point of Bitcoin’s success. This article proposes an Industry-4.0-compliant next-generation Bitcoin architecture by introducing a dynamic and sustainable block concept. Along with our modified knapsack algorithms, i.e., priority-based 0/1 knapsack and advanced-priority-based 0/1 knapsack, we can ensure a balanced transaction selection, quicker verification, higher transaction throughput, reduced carbon emission, and increased earnings for the miners. Moreover, with the addition of only one of our proposed sustainable blocks, we can cut down verification times by 50% and increase throughput by 39%. We can also reduce carbon emissions per transaction by 61.3%, which would help reduce Bitcoins’ large carbon footprint, enabling us to approach greener digital transactions.

With the continuous growth of internet users and media content, it is very hard to track down hateful speech in audio and video. Converting video or audio into text does not detect hate speech accurately as human sometimes uses hateful words as humorous or pleasant in sense and also uses different voice tones or show different action in the video. The state-of-the-art hate speech detection models were mostly developed on a single modality. In this research, a combined approach of multi-modal system has been proposed to detect hate speech from video contents by extracting feature images, feature values extracted from the audio, text and used machine learning and Natural language processing.

Edge-of-Things (EoT)-based healthcare services are forthcoming patient-care amenities related to autonomic and persuasive healthcare, where an EoT broker usually works as a middleman between the Healthcare Service Consumers (HSC) and Computing Service Providers (CSP). The computing service providers are the edge computing service providers (ECSP) and cloud computing service provider (CCSP). Sensor observations from a patient’s body area networks (BAN) and patients’ medical and genetic historical data are very sensitive and have a high degree of interdependency. It follows that EoT based patient monitoring systems or applications are tightly coupled and require obstinate synchronization. Therefore, this paper proposes a portfolio optimization solution for the selection of virtual machines (VMs) of edge and/or cloud computing service providers. The dynamic pricing for an EoT computation service is considered by the EoT broker for optimal VM provisioning in an EoT environment. The proposed portfolio optimization solution is compared with the traditional certainty equivalent approach. As the portfolio optimization is a centralized solution approach, this paper also proposes an alternating direction method of multipliers (ADMM) based distributed provisioning method for the healthcare data in the EoT computing environment. A comparative study shows the cost-effective provisioning for the healthcare data through portfolio optimization and ADMM methods over the traditional certainty equivalent and greedy approach, respectively.

In recent years, in order to provide a better quality of service (QoS) to Internet of Things (IoT) devices, the cloud computing paradigm has shifted toward the edge. However, the resource capacity (e.g., bandwidth) in fog network technology is limited and it is essential to efficiently bind the IoT applications with stringent QoS requirements with the available network infrastructure. In this paper, we formulate a joint user association and resource allocation problem in the downlink of the fog network, considering the evergrowing demand of QoS requirements imposed by the ultra-reliable low latency communications and enhanced mobile broadband services. First, we determine the priority of different QoS requirements of heterogeneous IoT applications at the fog network by enforcing the analytical framework using an analytic hierarchy process (AHP). Using the AHP, we then formulate a two-sided matching game to initiate stable association between the fog network infrastructure (i.e., fog devices) and IoT devices. Subsequently, we consider the externalities in the matching game that occurs due to job delay and solve the network resource allocation problem by applying the “best-fit” resource allocation strategy during matching. The simulation results illustrate the stability of the user association and efficiency of resource allocation with higher utility gain.

In recent years, Influence Maximization (IM) has gained great research interest in the field of social network research. The IM is a viral marketing based approach to find the influential users on the social networks. It determines a small seed set that can activate a maximum number of nodes in the network under some diffusion models such as Linear Threshold model or Independent Cascade model. However, previous works have not focused on the opportunity cost defined by the minimum number of nodes that must be motivated in order to activate the initial seed nodes. In this work, we have introduced a Reverse Influence Maximization (RIM) problem to estimate the opportunity cost. The RIM, working in opposite manner to IM, calculates the opportunity cost for viral marketing in the social networks. We have proposed the Extended Randomized Linear Threshold RIM (ERLT-RIM) model to solve the RIM problem. The ERLT-RIM is a Linear Threshold (LT)-based model which is an extension to the existing RLT-RIM model. We also have evaluated the performance of the algorithm using three real-world datasets. The result shows that the proposed model determines the optimal opportunity cost with time efficiency as compared to existing models.

Social networks have gained huge research interest, especially in viral marketing due to their rapid boom in the past years. It is very crucial to identify the influential users in the social networks for viral and target marketing. Influence maximization (IM) problem estimates such influential users in the social networks. With an initial seed set, the IM finds a maximum number of nodes that can be activated in the network under some diffusion models e.g. Linear Threshold model or Independent Cascade model. But previous works in this field have not studied about the minimum cost, termed as opportunity cost (OC), to motivate those seed nodes. In this work, we define a novel Reverse Influence Maximization (RIM) problem to determine the opportunity cost of influence maximization. Employing the influence propagation in opposite order, the RIM determines the minimum number of nodes that must be activated in order to motivate a set of target nodes. We propose Random RIM (R-RIM) and Randomized Linear Threshold RIM (RLT-RIM) models to tackle the RIM problem. We also perform a simulation to evaluate the performance of the algorithms using two real world datasets. The result shows that the proposed models determine the optimized opportunity cost with faster running time margin.

Modern day's User Equipments (UEs) are equipped with rich resources which encourage them to be used for more sophisticated applications. On the other hand, with these equipments in hand, users demand for high-quality video on the move is increasing day-by-day. Moreover, Information/Content Centric Networking (ICN/CCN) has changed the network dynamics by getting the desired contents regardless of the location. Unused memory in UEs can be used to cache the contents and provide it to the other nearby users on demand. In this paper, we propose to provide the requested video to users from other users cache, using D2D link, if it is present there. Our objective is to reduce the download delay for the users' requested video. We formulate the problem as a matching game in which the resources are assigned to the users in the uplink period. The UEs select the content node for D2D communication and the suitable channel. We have evaluated the proposed mechanism by implementing it in Matlab and have compared it with greedy approach and no D2D communication scheme. The experimental results show the effectiveness of our proposed mechanism.

Information Centric Networking (ICN) is envisioned to be the future Internet architecture and mobile access network e.g., Long Term Evolution (LTE), and 5G will be the major access networks. In this paper, we present a cache management and cooperative request forwarding schemes for Scalable Video Streaming (SVS) in Information Centric Networking (ICN) enabled mobile access networks. H.264/SVC encoded video is consisted a mandatory baselayer and multiple optional enhancement layers. Baselayer, which is enough to decode the video, though with the lowest quality, is needed by every user who want to watch the video while enhancement layers are used to improve the video quality. Only a subset of users download enhancement layers of the video. Therefore, caching the baselayer nearer to the users will increase their Quality of Experience. Furthermore, we introduce cooperative request forwarding for the baselayer of video to take more benefits from cache of neighboring base stations. We have intensively simulated our proposed schemes by extending chunk level simulator ccnsim which is developed over Omnet++. Our experimental results show that, cache hit rate can be improved significantly by adopting our proposed caching and Interest forwarding schemes.

The proliferation of the market in patient care services is attracting attention in the healthcare industry; however, a remote mental healthcare system is still unattainable. In this paper, an ambient intelligent system of in-home psychiatric care service for emergency psychiatry (EM-psychiatry) is proposed for the remote monitoring of psychiatric emergency patients. The emergency psychiatric states of patients are modeled as the states of the maximum-entropy Markov model (MEMM), in which sensor observations, psychiatric screening scores, and patients’ histories are considered as the observations of MEMM. A modified Viterbi, a machine-learning algorithm, is used to generate the most probable psychiatric state sequence based on such observations; then, from the most likely psychiatric state sequence, the emergency psychiatric state is predicted through the proposed algorithm. The ambient EM-psychiatry model is implemented and the performance of the proposed prediction model is analyzed using the receiver operator characteristics curves, which demonstrates that the use of the EM-psychiatric screening questionnaire with biosensor observations enhances the prediction accuracy.

The revolutionized vision of IoT has united heterogeneous devices to foster the systems of cohesive intelligent things. In addition, Fog computing has also envisioned a new form of cloud computing paradigm. Therefore, Fog provides edge computing to such IoT devices with varied capabilities and resources. However, a balanced and efficient pairing or matching strategy for edge IoT nodes is crucial to achieve the user requisite. Hence, this paper addresses the utility based matching or pairing problem within the same domain of IoT nodes by using Irving's matching algorithm under the node specified preferences to endure a stable IoT node pairing. We studied the performance of the proposed matching algorithm through simulation. The simulation results show the higher utility gain of the node pairs through refined matching algorithm over greedy approach.

We take the first step to study the price competition in a heterogeneous market cloud computing formed by public provider and cloud broker, all of which are also known as cloud service providers. We formulate a price competition between cloud broker and public provider as a two-stage non-cooperative game. In stage one, where cloud service providers set their service prices to maximize their revenue, we use the Nash equilibrium concept to study the equilibria for the price setting game. Cloud users can select the services (from the cloud broker or public provider) that provide them the best payoff in terms of performance (i.e., delay) and price. To that end, cloud users can adapt their service selection behavior by observing the variations in price and quality of service offered by the different cloud service providers. For the service selection game of cloud users in stage two, we use the evolutionary game model to study the evolution and the dynamic behavior of cloud users. Furthermore, the Wardrop equilibrium and replicator dynamics is applied to determine the equilibrium and its convergence properties of the service selection game. Numerical results illustrate that our game model captures the main factors behind the heterogeneous market cloud pricing and service selection, thus represents a promising framework for the design and understanding of the heterogeneous market cloud computing.

According to increasing performance of mobile devices, like smart phone, tablet PC and etc, and diffusing network infrastructures, like LTE, WiFi and etc, various types of content delivery services based on PC services can serve into mobile devices using cloud. In this paper we proposed content delivery framework with SDN (Software Defined Networking) and CCN (Content Centric Networking) to improve content delivery QoS in mobile cloud environment. Additionally to serve autonomic optimal services, we proposed reinforcement learning based context-aware content delivery scheme. Using our framework, we can guarantee QoS to provide context-aware content delivery scheme.

Coexistence of different wireless networks is a common phenomenon in today's smart communication infrastructure. Now, the big issue is to explore benefits from the heterogeneous nature of communication technology. Load balancing among the heterogeneous wireless networks is the primary goal of this paper. Load balancing without considering Quality of Service (QoS) merely inadequate in convergence of resource utilization and grade of service. So, this paper proposed a load balancing algorithm with QoS provisioning. This paper is based on a semi-distributed load balancing architecture. Firstly, IP-flow dividing ratio based soft load balancing approach is discussed for high speed features of next generation wireless networks. Secondly, an admission control function of QoS requirements is developed. Thirdly, a joint optimization function is derived and a load balancing algorithm is proposed by using the cost function. Finally, simulation results are presented for performance appraisal.