Explore the vast range of titles available.

This paper introduces a new image-based handwritten historical digit dataset named Arkiv Digital Sweden (ARDIS). The images in ARDIS dataset are extracted from 15,000 Swedish church records which were written by different priests with various handwriting styles in the nineteenth and twentieth centuries. The constructed dataset consists of three single-digit datasets and one-digit string dataset. The digit string dataset includes 10,000 samples in red–green–blue color space, whereas the other datasets contain 7600 single-digit images in different color spaces. An extensive analysis of machine learning methods on several digit datasets is carried out. Additionally, correlation between ARDIS and existing digit datasets Modified National Institute of Standards and Technology (MNIST) and US Postal Service (USPS) is investigated. Experimental results show that machine learning algorithms, including deep learning methods, provide low recognition accuracy as they face difficulties when trained on existing datasets and tested on ARDIS dataset. Accordingly, convolutional neural network trained on MNIST and USPS and tested on ARDIS provide the highest accuracies 58.80 % and 35.44 % , respectively. Consequently, the results reveal that machine learning methods trained on existing datasets can have difficulties to recognize digits effectively on our dataset which proves that ARDIS dataset has unique characteristics. This dataset is publicly available for the research community to further advance handwritten digit recognition algorithms.

Recognizing human activities using automated methods has emerged recently as a pivotal research theme for security-related applications. In this research paper, an optical flow descriptor is proposed for the recognition of human actions by considering only features derived from the motion. The signature for the human action is composed as a histogram containing kinematic features which include the local and global traits. Experimental results performed on the Weizmann and UCF101 databases confirmed the potentials of the proposed approach with attained classification rates of 98.76% and 70%, respectively, to distinguish between different human actions. For comparative and performance analysis, different types of classifiers including Knn, decision tree, SVM and deep learning are applied to the proposed descriptors. Further analysis is performed to assess the proposed descriptors under different resolutions and frame rates. The obtained results are in alignment with the early psychological studies reporting that human motion is adequate for the perception of human activities.

A number of full reference and reduced reference methods have been proposed in order to estimate the perceived visual quality of 3D meshes. However, in most practical situations, there is a limited access to the information related to the reference and the distortion type. For these reasons, the development of a no-reference mesh visual quality (MVQ) approach is a critical issue, and more emphasis needs to be devoted to blind methods. In this work, we propose a no-reference convolutional neural network (CNN) framework to estimate the perceived visual quality of 3D meshes. The method is called SCNN-BMQA (3D visual saliency and CNN for blind mesh quality assessment). The main contribution is the usage of a CNN and 3D visual saliency to estimate the perceived visual quality of distorted meshes. To do so, the CNN architecture is fed by small patches selected carefully according to their level of saliency. First, the visual saliency of the 3D mesh is computed. Afterward, we render 2D projections from the 3D mesh and its corresponding 3D saliency map. Then the obtained views are split into 2D small patches that pass through a saliency filter in order to select the most relevant patches. Finally, a CNN is used for the feature learning and the quality score estimation. Extensive experiments are conducted on four prominent MVQ assessment databases, including several tests to study the effect of the CNN parameters, the effect of visual saliency and comparison with existing methods. Results show that the trained CNN achieves good rates in terms of correlation with human judgment and outperforms the most effective state-of-the-art methods.

The article deals with the problems of motion detection, object recognition, and scene description using deep learning in the framework of granular computing and Z-numbers. Since deep learning is computationally intensive, whereas granular computing, on the other hand, leads to computation gain, a judicious integration of their merits is made so as to make the learning mechanism computationally efficient. Further, it is shown how the concept of z-numbers can be used to quantify the abstraction of semantic information in interpreting a scene, where subjectivity is of major concern, through recognition of its constituting objects. The system, thus developed, involves recognition of both static objects in the background and moving objects in foreground separately. Rough set theoretic granular computing is adopted where rough lower and upper approximations are used in defining object and background models. During deep learning, instead of scanning the entire image pixel by pixel in the convolution layer, we scan only the representative pixel of each granule. This results in a significant gain in computation time. Arbitrary-shaped and sized granules, as expected, perform better than regular-shaped rectangular granules or fixed-sized granules. The method of tracking is able to deal efficiently with various challenging cases, e.g., tracking partially overlapped objects and suddenly appeared objects. Overall, the granulated system shows a balanced trade-off between speed and accuracy as compared to pixel level learning in tracking and recognition. The concept of using Z-numbers, in providing a granulated linguistic description of a scene, is unique. This gives a more natural interpretation of object recognition in terms of certainty toward scene understanding.

Mining association patterns from a time-stamped temporal database distributed over finite time slots is implicitly associated with task of scanning the input database. Finding supports of itemsets requires scanning the input database. Database scan can be performed in either snapshot or lattice-based approach. Sequential and SPAMINE methods for similarity-profiled association pattern mining originally proposed by Jin Soung Yoo and Sashi Sekhar are based on the snapshot database scan and lattice scan, respectively. Snapshot database scan involves scanning multi-time slot database time slot by time slot. The major limitation of Sequential method is the requirement to retain original temporal database in the disk for finding itemset support computations. In this paper, a novel multi-tree structure called VRKSHA is proposed that eliminates the need to store the original temporal database in the memory and also eliminates the need to retain database in memory. The basic idea is to generate a compressed time-stamped temporal tree and use this multi-tree structure to obtain true supports of temporal itemsets for a given time slot. Discovery of similar temporal itemsets is based on finding distance between temporal itemset and reference w.r.t each time slot and validating whether the computed distance satisfies specified user dissimilarity threshold. A pattern is pruned if the dissimilarity condition fails at any given time slot well before computing true support of itemset w.r.t all time slots. The advantage of proposed Sequential approach is from the fact that it is a single database scan approach excluding the initial database scan performed for computing true supports of singleton items. VRKSHA overcomes the major limitation of retaining database in memory that is required by SPAMINE, G-SPAMINE, MASTER algorithms. Experiment results prove that computational time and memory consumed by VRKSHA are significantly very much better than by approaches such as Naïve, Sequential, SPAMINE, and G-SPAMINE. To the best of our survey and knowledge, VRKSHA is the pioneering work to introduce and propose a compressed tree-based data structure for mining similarity-profiled temporal association patterns in the area of time-profiled temporal association mining.

An intelligent employment rate prediction model based on a neural computing framework and human–computer interaction platform is demonstrated in this manuscript. Predictive analytics is the future of things, and its significance is manifested in two main aspects: understanding the future so that people can prepare for its arrival, and predicting the current decision so that people can understand the possible consequences, and by the consequences of the analysis to determine the current decision, and strive to make the current decision. However, there are lots of challenges for the prediction tasks. The novelty of this research is mainly concentrated on two major aspects: (1) the neural network model is optimized and enhanced. The proposed nerve tree network model is essentially based on a tree-structured code for a multi-layered feed-forward sparse neural network; with the tree-structured code, the nerve tree network model does not require interconversion between its genotype and phenotype in the coding and decoding operations, and also effectively reduces the computing time. (2) The human–computer interaction is integrated to construct a user-friendly system. In interactive technology, the interactive contact surface and the model, interactive methods and social acceptance have also given rise to many questions that must be solved and problems that require further research and technological innovation. Through numerical verification, the performance of the proposed framework is validated, and the simulation proves the overall performance of the proposed model. Compared with other models, the proposed algorithms can achieve higher prediction accuracy.

Crowd density estimation is a challenging research problem in computer vision and has many applications in commercial and defense sectors. Various crowd density estimation methods have been proposed by researchers in the past, but there is an utmost need for accurate, robust and efficient crowd density estimation techniques for its practical implementation. In this paper, we propose a fine-tuned and computationally economical, ensemble regression-based machine learning model for crowd density estimation. The WorldExpo’10 dataset has been used for experimental analysis and model performance evaluation. We extract variety of features in texture - based features such as gray-level co-occurrence matrix, local binary pattern and histogram of oriented gradients, structure - based features such as perimeter pixel and the orientation of pixels, and segment-based handcrafted features from each patch of the image and use an optimum combination of these features as input to the regression model. To achieve optimized memory utilization and faster speed, principal component analysis is employed to reduce the dimensions of the lengthy feature vector. Extensive experiments on different fronts ranging from the model hyperparameter optimization, features optimization and features selection were conducted, and at each step, we selected the most favorable results as input to the optimized model. The performance of the model is evaluated based on two popular metrics, i.e., mean absolute error and mean squared error. The comparative analysis shows that the proposed system outperforms the former methods tested on the WorldExpo’10 dataset.

In our modern life, persons and institutions alike are rapidly embracing the shift toward communication via the Internet. As these entities adopt a faster and efficient communication protocol, information security techniques such as steganography and cryptography become powerful and necessary tools for conducting secure and secrecy communications. Currently, several steganography techniques have been developed, and the least significant bit (LSB) is one of these techniques which is a popular type of steganographic algorithms in the spatial domain. Indeed, as any other existing techniques, the selection of positions for data embedding within a cover signal mainly depends on a pseudorandom number generator without considering the relationship between the LSBs of the cover signal and the embedded data. In this paper and for best pixels’ positions adjustment, in which the visual distortion of the stego-image, as well as the embedding changes, becomes optimum, we propose two new position selection scenarios of LSBs-based steganography. Our new works are to improve the embedding efficiency, that is to say, select the suitable cover image pixels’ values that optimize the expected number of modifications per pixel and the visual distortion.

There has recently been a major advance with respect to how information fusion is performed. Information fusion has gone from being conceived as a purely hierarchical procedure, as is the case of traditional military applications, to now being regarded collaboratively, as holonic fusion, which is better suited for civil applications and edge organizations. The above paradigm shift is being boosted as information fusion gains ground in different non-military areas, and human–computer and machine–machine communications, where holarchies, which are more flexible structures than ordinary, static hierarchies, become more widespread. This paper focuses on showing how holonic structures tend to be generated when there are constraints on resources (energy, available messages, time, etc.) for interactions based on a set of fully intercommunicating elements (peers) whose components fuse information as a means of optimizing the impact of vagueness and uncertainty present message exchanges. Holon formation is studied generically based on a multiagent system model, and an example of its possible operation is shown. Holonic structures have a series of advantages, such as adaptability, to sudden changes in the environment or its composition, are somewhat autonomous and are capable of cooperating in order to achieve a common goal. This can be useful when the shortage of resources prevents communications or when the system components start to fail.

The improvement of human quality of life is inseparable from the support of information technology, and the development of information technology has made human life more convenient. The current era is the information age, and the level of informatization has gradually become one of the important indicators to measure the comprehensive level of a country. The emergence of the Internet of Things has led to rapid development of technologies such as data perception, wireless data transmission, and intelligent information processing. With the increasing use of information transmission, people gradually realize the impact of security issues on themselves and society. In this paper, a smart IoT information transmission and security optimization model based on chaotic neural computing model is proposed. Simulation and analysis show that the proposed algorithm can ensure the availability and confidentiality of data at the same time.