首页期刊导航|Multimedia tools and applications
期刊信息/Journal information
Multimedia tools and applications
Kluwer Academic Publishers
Multimedia tools and applications

Kluwer Academic Publishers

1380-7501

Multimedia tools and applications/Journal Multimedia tools and applicationsSCIISTPEIAHCI
正式出版
收录年代

    CatRevenge: towards efective revenge text detection in online social media with paragraph embedding and CATBoost

    Sayani GhosalAmita Jain
    89607-89633页
    查看更多>>摘要：Huge amount of internet data are produced and consumed by internet users, where most of the data are in natural language and they express their feelings, emotions and thoughts on social media. It is the responsibility of the social media provider to provide healthy com- munication system among users. It is very challenging job to detect revenge from the social media text due to long sentences where semantic relation dissolves between tokens. Due to that, the social media providers did not provide any attention towards identifying the users spreading revenge. This article propose a novel model named as CatRevenge which identi- fes both active and passive revenge. This model preprocess with Slangzy internet slang meaning dictionary to detect revenge text more efciently. CatRevenge assigns impact weight on each of parts of speech in the sentences based on its relevance and TF-IDF score of the words. The novel CatRevenge model also considers the paragraph embedding model for contextual semantic analysis of revenge text. In addition, this research applies gradi- ent boosting CATBoost classifer with categorical features to reduce model overftting. This feature ranking method can able to reduce the dimensionality of data by ranking the most signifcant feature. This research considers the revenge posts English language dataset from the Reddit social media where it evaluated with binary and multiclass classifcation. Results demonstrate achievable performance with a 6-10% increase in binary and a 2.5 -5% increase in multiclass with weighted F1 metric.
      原文链接:
    • NETL
    • NSTL

    Ensemble of deep learning techniques to human activity recognition using smart phone signals

    Soodabeh ImanzadehJafar TanhaMahdi Jalili
    89635-89664页
    查看更多>>摘要：Human Activity Recognition (HAR) has become a signifcant area of study in the felds of health, human behavior analysis, the Internet of Things, and human-machine interaction in recent years. Smartphones are a popular choice for HAR as they are common devices used in daily life. However, most available HAR datasets are gathered in laboratory settings, which do not refect real-world scenarios. To address this issue, a real-world dataset using smartphone inertial sensors, involving 62 individuals, is collected. The collected dataset is noisy, small, and has variable frequency. On the other hand, in the context of HAR, algo- rithms face additional challenges due to intra-class diversity (which refers to diferences in the characteristics of performing an activity by diferent people or by the same individual under diferent conditions) and inter-class similarity (which refers to diferent activities that are highly similar). Consequently, it is essential to extract features accurately from the dataset. Ensemble learning, which combines multiple models, is an efective approach to improve generalization performance. In this paper, a weighted ensemble of hybrid deep models for HAR using smartphone sensors is proposed. The proposed ensemble approach demonstrates superior performance compared to current methods, achieving impressive results across multiple evaluation metrics. Specifcally, the experimental analysis demon- strates an accuracy of 97.15%, precision of 96.41%, recall of 95.62%, and an F1-score of 96.01%. These results demonstrate the efectiveness of our ensemble approach in address- ing the challenges of HAR in real-world scenarios.
      原文链接:
    • NETL
    • NSTL

    Machine learning‑based early detection of diabetes risk factors for improved health management

    Praveena NuthakkiT. Pavan Kumar
    89665-89680页
    查看更多>>摘要：This research requires to improve the accuracy of early diabetic forecasting in a human body or patient by applying diverse machine learning approaches. Approaching to creation of machine learning models by using patient datasets to produce predictions with improved accuracy. This work will use machine learning classifcation and ensemble approaches, such as Random Forest (RF), Gradient Boosting (GB), Decision Tree (DT), K-nearest neighbour (KNN), Logistic Regression (LR), and Support Vector Machine (SVM), on a dataset to predict diabetes. The accuracy of each model difers in comparison to other models. This work demonstrates the model's capability by providing an accurate or greater accuracy. This research paper reported diferent performance metrics like precision, recall, accuracy, F1 score, and sensitivity for various machine learning algorithms. Final experi- mental results indicate that the Random Forest classifer outperforms other methods.
      原文链接:
    • NETL
    • NSTL

    A novel secure communication system using Dufng's chaotic model

    Maryan Mohamed ManhilRaied K. Jamal
    89681-89694页
    查看更多>>摘要：In this work the chaotic Dufng system was used in secure communications applications. The shift-keying scheme was used to modulate the diferent signals such as; OVEN and the sinusoidal signal. It has been proven that the chaotic carriers allow the process of modu- lating and demodulating signals easily and with high efciency, and this is coupled with the coupling factor between the receiving and transmittance units. Due to the absence of identical chaotic systems in time series behavior, it is assumed that there are two diferent chaotic systems in the initial conditions, one of which is called the transmitting unit and the second is called the receiving unit, which are as follows: xi1, yi1, xi2, yi2, where the two systems were coupling to an external stimulus called the coupling factor. The coupling factor k was changed from 0 to 100 a.u. to obtain full synchronization at this value for the sinusoidal signal and OVEN. In addition, there are limitations for the frequency and ampli- tude values of the sinusoidal signal on the signal transmission secret, and limitations for the input values for the OVEN signal. After experimenting with many values of the cou- pling factor, it was found that the two units have complete synchronization when the value of k=100. Where the error function becomes zero and falls within±0.001. It was observed that at amplitude and frequency A=0.1, f=0.01. the transmitted signal is clear and there is no secrecy in transmitting information, while at amplitude A=0.1, f=0.34 the signal is implicitly hidden within the chaos signal.
      原文链接:
    • NETL
    • NSTL

    An optimized ensemble model based on cuckoo search with Levy Flight for automated gastrointestinal disease detection

    Zafran WaheedJinsong Gui
    89695-89722页
    查看更多>>摘要：Accurate detection of gastrointestinal (GI) diseases is critical for efective medical inter- vention. Existing methods often lack accuracy and efciency, emphasizing the need for more advanced approaches. The complexity and diversity of medical image data, such as those found in GI diseases, can pose challenges for a single model to comprehensively represent all essential features. In such scenarios, an ensemble learning approach becomes important. In this paper, we propose an innovative ensemble learning approach for GI dis- ease prediction. We leverage the power of three transfer learning models, DenseNet169, InceptionV3, and MobileNet, as base learners along with additional layers to efectively learn data-specifc features. We implement a weighted averaging ensemble strategy to merge predictions from individual base models and fne-tune the weights using the cuckoo search (CS) with levy fight algorithm. This approach results in more accurate predictions compared to individual models, leveraging the diverse strengths of the base learners for enhanced performance in GI disease prediction. This study is notably the pioneer in intro- ducing a metaheuristics-based optimized model for the detection of GI diseases. We assess the presented model using a publicly accessible endoscopic image dataset that consists of 6,000 images. The results demonstrate exceptional predictive accuracy, with the ensem- ble achieving an outstanding accuracy of 99.75%. Through Grad-CAM analysis, we gain valuable insights into the decision-making process of the individual base models, enabling us to identify areas of strength and improvement. Our proposed ensemble model outper- forms traditional weight assignment methods and existing state-of-the-art methods, show- casing its superiority in GI disease prediction. Our approach has transformative potential in medical image analysis, promising enhanced patient care and diagnostic accuracy in gastroenterology.
      原文链接:
    • NETL
    • NSTL

    An oceanographic data collection scheme using hybrid optimization for leakage detection during oil mining in mobility assisted UWSN

    Monika ChoudharyNitin GoyalDeepali GuptaBhanu Sharma...
    89723-89741页
    查看更多>>摘要：Data acquisition is the process of collecting, measuring and analysing information using standardised, validated techniques for application-specifc tasks. In mobility-assisted underwater wireless sensor networks (UWSNs), where nodes are not fxed due to water current of 3 m/sec, data collection becomes an arduous task. There are few works that pro- vide a mobile sink with optimised data transmission path planning and scheduling. These systems do not transmit the data fast enough to provide real-time data transmission as these methods do not consider the bufer occupancy rate and latency in data acquisition. In this paper, a stimulating transmission path planning technique for mobile sinks using the hybrid Grey Wolf Optimizer Whale Optimization Algorithm (GWOWOA) is proposed. In contrast to other optimization techniques, this hybrid technique includes a number of update pro- cesses such as random position update, prey search by the Grey Wolf Optimizer (GWO) and prey search by the Whale Optimization Algorithm (WOA). In this paper, the ftness function is calculated in terms of distance to the mobile sink, bufer occupancy rate, energy level and data acquisition latency. The use of these variables makes the proposed technique innovative. To prove the efciency of the proposed system, GWOWOA is compared with existing systems. The simulation results show that the proposed system increases the resid- ual energy and accuracy of the collected data and minimises the delay.
      原文链接:
    • NETL
    • NSTL

    A systematic review of virtual 3D reconstructions of Cultural Heritage in immersive Virtual Reality

    Bruno Rodriguez‑GarciaHenar Guillen‑SanzDavid ChecaAndres Bustillo...
    89743-89793页
    查看更多>>摘要：Immersive Virtual Reality (iVR) devices are increasingly afordable and accessible to consumers. The widespread adoption of this technology for professional training is now fnding its way into various other felds. One feld that is gaining signifcant popularity is Cultural Heritage (CH), where iVR enables the reconstruction and exploration of lost herit- age. However, an up-to-date systematic review of iVR within this feld will be of great ben- eft. Hence, the present review of 94 papers published between 2013 and 2022 that follows PRISMA methodology on virtual reconstruction of CH for iVR. The aim is to identify the key factors behind the development of these applications and their standards. To do so, a statistical analysis on the following topics was performed: (1) nationality, publication date, and article type; (2) heritage type and its current state of preservation; (3) the area of fnal application and the features of the reconstructions; (4) the characteristics of the iVR expe- rience; and (5) the assessment of the iVR applications. Finally, a roadmap of best practices is outlined for the virtual reconstruction of CH using iVR and some of the most promising future research lines are outlined.
      原文链接:
    • NETL
    • NSTL

    Euclidean embedding with preference relation for recommender systems

    V Ramanjaneyulu YannamJitendra KumarKorra Sathya BabuBidyut Kumar Patra...
    89795-89815页
    查看更多>>摘要：Recommender systems (RS) help users pick the relevant items among numerous items that are available on the internet. The items may be movies, food, books, etc. The Recommender systems utilize the data that is fetched from the users to generate recommendations. Usually, these ratings may be explicit or implicit. Explicit ratings are absolute ratings that are gener- ally in the range of 1 to 5. While implicit ratings are derived from information like purchase history, click-through rate, viewing history, etc. Preference relations are an alternative way to represent the users' interest in the items. Few recent research works show that preference relations yield better results compared to absolute ratings. Besides, in RS, the latent fac- tor models like Matrix Factorization (MF) give accurate results especially when the data is sparse. Euclidean Embedding (EE) is an alternative latent factor model that yields similar results as MF. In this work, we propose a Euclidean embedding with preference relation for the recommender system. Instead of using the inner product of items and users' latent factors, Euclidean distances between them are used to predict the rating. Preference Relations with Matrix Factorization (MFPR) produced better recommendations compared to that of tradi- tional matrix factorization. We present a collaborative model termed EEPR in this work. The proposed framework is implemented and tested on two real-world datasets, MovieLens-100K and Netflix-1M to demonstrate the effectiveness of the proposed method. We utilize popular evaluation metric for recommender systems as precision@K. The experimental outcomes show that the proposed model outperforms certain state-of-the-art existing models such as MF, EE, and MFPR.
      原文链接:
    • NETL
    • NSTL

    Swin-TransUper: Swin Transformer-based UperNet for medical image segmentation

    Jianjian YinYi ChenChengyu LiZhichao Zheng...
    89817-89836页
    查看更多>>摘要：Convolutional Neural Network-based UNet and its variants have shown remarkable per- formance in medical image segmentation. However, these methods can only capture local features without spatial correlations and are incapable of global modeling. Previous studies prove that local and global features are critical in computer vision. Therefore, based on the abovementioned considerations, this paper proposes a pure Transformer model named Swin- TransUper. Firstly, we explore extending UperNet by incorporating the hierarchical Swin Transformer with shifted windows, thereby enhancing the global modeling capability of the model. Secondly, we introduce an SPPM (Swin Pyramid Pooling Module) to conduct multi- scale feature mining on the deepest features generated by the encoder, fully considering the semantic information of the deepest features. Finally, the multi-scale attention module aggre- gates the multi-scale feature information to obtain a more refined feature map. Our method achieves the state-of-the-art performance of 80.08%, 90.25%, and 90.62% on the Synapse multi-organ segmentation, ISIC2017, and ACDC datasets based on the DSC (Dice Similarity Coefficient) metric. At the same time, experimental results on the ISIC2017 dataset show that Swin-TransUper achieves the best performance on Sensitivity and Accuracy metrics of 91.20% and 96.44%, respectively.
      原文链接:
    • NETL
    • NSTL

    Grading the severity of diabetic retinopathy using an ensemble of self‑supervised pre‑trained convolutional neural networks: ESSP‑CNNs

    Saeed ParsaToktam Khatibi
    89837-89870页
    查看更多>>摘要：Diabetic retinopathy (DR) is a common eye disorder that can lead to vision problems and blindness, necessitating accurate grading for efective treatment. While various artifcial intelligence (AI) systems have been developed, surpassing human analysis in detecting DR, deep neural networks require large annotated datasets to learn the complex patterns and relationships necessary for grading, which are often limited in availability, to learn the intricate patterns and relationships required for accurate grading. However, such data- sets are often limited in availability, requiring signifcant investments of human resources and time for the labeling process. To address these challenges, we propose ESSP-CNNs, a framework that harnesses popular CNN architectures (VGGNet, AlexNet, and ResNet). Our approach employs self-supervised learning, specifcally the Bootstrap Your Own Latent (BYOL) technique, to pre-train neural networks on a vast unlabeled dataset. Additionally, we employ deep ensemble learning to construct a robust model for DR grading. Our meth- odology encompasses three main components: preprocessing fundus images, BYOL-based pre-training, and ensemble model construction. We conduct experiments and comparisons using the EyePACS and IDRiD datasets, with BYOL pre-training on EyePACS to enable the CNN models to acquire meaningful representations of fundus images, while IDRiD is used for severity grading. The performance of the proposed framework is further con- frmed through thorough validation using the Messidor dataset. Through extensive experi- mentation on the IDRiD and Messidor datasets, ESSP-CNNs achieve notable accuracies of 71.84% and 75.42%, specifcities of 88.76% and 87.13% along with AUC of 86.02% and 86.54%, respectively. The experimental results validate the efectiveness of our methodol- ogy in grading the severity of DR, with the ensemble model built from pre-trained CNNs yielding promising outcomes. Moreover, we compare our methodology against other state- of-the-art methods in DR grading, and our results demonstrate its satisfactory performance, surpassing previous alternatives in accurately assessing DR severity.
      原文链接:
    • NETL
    • NSTL