Search

The paper deals with the problem of acoustic events classification which is actively applied to the problems of a safe city, smart home, IoT devices, and for the detection of industrial accidence. A solution to improve the accuracy of classifiers without changing their structure and collecting additional data is proposed. The main data source for the experiments was the TUT Urban Acoustic Scenes 2018, Development Dataset. The paper presents the way to increase the accuracy of audio event classification by using the FN-corrector. The FN-corrector is a linear two-stage classifier performing the transformation of the feature space into a linearly separable space and the linear separation of one class from another. If a corrector is applied, the responses of the original classifier generate four classes: positive (P), negative (N), false positive (FP), and false negative (FN). As a result, it becomes possible to train two types of correctors: the FP-corrector separating positive and false positive classifier responses, and the FN-corrector separating negative and false negative classifier responses. In the experiments, the VGGish convolutional neural network was used as the initial classifier. The audio signal is converted into a spectrogram and is fed to the input of the neural network which forms the spectrogram feature description and performs a classification. As an example, two ”confused“ classes are selected to demonstrate the increase in classification accuracy. Using the feature description of audio recordings of these classes, an FN-corrector was built, trained and connected to the original classifier. The response from the classifier, as well as the feature description, has been passed to the corrector input. Next, the corrector translated the feature space into a new basis (into a linearly separable space) and classified the classifier answer responding to the question whether the original classifier makes a mistake on such a feature vector or not. If the original classifier made a mistake, then his answer is changed by the corrector to the opposite, otherwise the answer remains the same. The results of the experiments demonstrated a decrease in the level of class confusion and, accordingly, an increase in the accuracy of the original classifier without changing its structure and without collecting an additional data set. The results obtained can be used on IoT devices that have significant limitations on the size of the models used, as well as in solving the problems of domain adaptation which is relevant in audio analytics.

Problems of Wireless Sensor Networks (WSN) are associated with a significant increase in the number of devices on these networks. In this regard, the requirements for the protection and the security of WSN from external influences are increasing significantly. WSN security problems are solved by solving the problem of optimal path routing, energy conservation, and so on. This paper proposes a hybrid model of an efficient packet routing and delivery system to prevent Black-hole attacks. This type of attack is considered the most common on the network due to its unique characteristics. To detect such attacks, a deep learning model using a Convolutional Neural Network (CNN) is proposed. The learning algorithm must be reliable and trustworthy so that attack analysis can be considered at different levels to study the intelligent behavior of network attacks. The paper considers the problem of finding the optimal shortest path using Deep Q-Learning and convolutional neural networks to perform efficient routing and delivery of packets in a safer way. As a result of simulation, the achieved accuracy reached 98.57 %.

The development of the medical imaging field is becoming a significant challenge due to the growing need for automated, fast, and efficient diagnostics. Traditionally, blood cells are counted by using a hemocytometer along with other laboratory equipment and chemicals, which is a laborious task. The paper investigates the application of machine learning methods to the identification and classification of blood cells, which allow increasing the recognition rate without deteriorating quality. A comparative analysis of methods for solving the problem of recognizing blood cells based on artificial intelligence approaches is carried out. The paper uses support vector machine, k-nearest neighbors’ algorithm, deep learning (convolutional neural network), and forward propagation neural network. A set of images with cell samples was selected as the initial data for comparison. A comparative analysis of the quality of the considered algorithms was performed on a set of training data with more than 3000 images. It is shown that a program that implements artificial intelligence methods provides a cell recognition time within 4-6 seconds when using an office personal computer, which is significantly less than the time spent by medical workers on one study of a biomaterial. The implementation of the presented results makes it possible to automate the process of studying a biomaterial, reduce the time for conducting and obtaining the result of the analysis of whole blood cells (identification and counting), lessen the influence of operator errors on the result, unload computing resources, thereby increases the efficiency of digital medicine.

The paper proposes a method of visual analysis for automatic speech recognition of the vehicle driver. Speech recognition in acoustically noisy conditions is one of big challenges of artificial intelligence. The problem of effective automatic lip-reading in vehicle environment has not yet been resolved due to the presence of various kinds of interference (frequent turns of driver's head, vibration, varying lighting conditions, etc.). In addition, the problem is aggravated by the lack of available databases on this topic. A MediaPipe Face Mesh is used to find and extract the region-of-interest (ROI). We have developed End-to-End neural network architecture for the analysis of visual speech. Visual features are extracted from a single image using a convolutional neural network (CNN) in conjunction with a fully connected layer. The extracted features are input to a Long Short-Term Memory (LSTM) neural network. Due to a small amount of training data we proposed that a Transfer Learning method should be applied. Experiments on visual analysis and speech recognition present great opportunities for solving the problem of automatic lip-reading. The experiments were performed on an in-house multi-speaker audio-visual dataset RUSAVIC. The maximum recognition accuracy of 62 commands is 64.09 %. The results can be used in various automatic speech recognition systems, especially in acoustically noisy conditions (high speed, open windows or a sunroof in a vehicle, backgoround music, poor noise insulation, etc.) on the road.

The importance of wearing a mask in public places came to light when the COVID-19 pandemic has started due to the coronavirus. To strictly control the spread of the virus, wearing a mask is mandatory to avoid getting the virus through others or spreading the virus to others if we are carrying it. Since it’s not possible to check each individual in public places whether he/she is wearing a mask, this paper proposed a face mask detection using Deep Learning (DL) and Convolutional Neural Network (CNN) techniques. A cloud-based approach that adopted DL is used to identify the persons violating the rules. The dataset used in the work is collected from various studies, such as Prajnasb/observations and Kaggle’s Face Mask Detection Dataset that contains images of people wearing and not wearing masks. The faces in the images will be detected and cropped with the help of a trained face detector which will be used for checking whether the face in the image is wearing a mask or not. Face mask detection is done with the help of CNN. The input image is fed into the CNN and the output is binary format, whether person wearing or not wearing a mask. The work uses Max Pooling and Average Pooling layers of CNN. The outcome of the work shows that the proposed method achieves 98 % of accuracy using Max Pooling which is better than the currently available works.

The relevance of the tasks of detecting and recognizing objects in images and their sequences has only increased over the years. Over the past few decades, a huge number of approaches and methods for detecting both anomalies, that is, image areas whose characteristics differ from the predicted ones, and objects of interest, about the properties of which there is a priori information, up to the library of standards, have been proposed. In this work, an attempt is made to systematically analyze trends in the development of approaches and detection methods, reasons behind these developments, as well as metrics designed to assess the quality and reliability of object detection. Detection techniques based on mathematical models of images are considered. At the same time, special attention is paid to the approaches based on models of random fields and likelihood ratios. The development of convolutional neural networks intended for solving the recognition problems is analyzed, including a number of pre-trained architectures that provide high efficiency in solving this problem. Rather than using mathematical models, such architectures are trained using libraries of real images. Among the characteristics of the detection quality assessment, probabilities of errors of the first and second kind, precision and recall of detection, intersection by union, and interpolated average precision are considered. The paper also presents typical tests that are used to compare various neural network algorithms.

Face detection is one of the most popular computer vision tasks. There are a lot of face detection approaches proposed including different CNN-based techniques, but the problem of optimal balancing between detection quality and computational speed is still relevant. In this paper we propose new CNN-based solution for face detection called FaceDetectNet. Our CNN architecture is based on ideas of YOLO/DetectNet and GoogleNet architecture supported with some new tools and implementation details created especially for our face detection application. We propose: original iterative proposal clustering (IPC) algorithm for aggregation of output face proposals formed by CNN and the 2-level “weak pyramid” providing better detection quality on the testing sets containing both small and huge images. Our face detection approach is close to previously proposed SSD-based face detection, but the principal difference is that we use the deep features of top hidden CNN layer for forming the face proposals of any size. Thus we utilize the global semantic and context information for improving the detection quality for small faces. Our FaceDetectNet is trained and tested on the most challenging WIDER FACE detection benchmark. Our algorithm achieves the average precision (AP) 0.69 on the WIDER FACE hard level, and thus outperforms all competitive detectors on the Hard level besides the HR state-of-the-art solution. Note that HR solution is based on essentially deeper and slower CNN, while our FaceDetectNet can work in real-time on the NVIDIA GeForce 1080 GPU. On the other hand, SSD-based face detector with comparable CNN parameters provides AP 0.625 only on the WIDER FACE hard level. So, our approach provides the best quality with reasonable computational speed.

This paper is an analytical review of an article by two researchers - Zeinab Ghasemi Tari (University of Tehran, Iran) and Zahra Emamzadeh (University of Canterbury, New Zealand), published under the title "An analysis of media reports during the 2016 U.S. presidential election: a Thematic comparison of CNN news and Donald Trump's tweets" in the journal of politics and law in 2018. The subject of study of the authors is the media confrontation between Donald Trump and CNN during the presidential elections in the United States. The main advantage of the considered foreign article is a detailed thematic analysis of media messages (Trump's tweets and CCN videos), which were exchanged by information opponents in the midst of the election campaign in 2016.

Subject of Research. The present paper is devoted to the attacks detection problem on voice biometric systems (spoofing-attacks) in telephone channel. Nowadays, spoofing detection is under the high interest in the field of voice speaker authentication. The results of the Automatic Speaker Verification Spoofing and Countermeasures Challenge in 2015 and 2017 dedicated to isolated task of spoofing detection confirmed the high perspectives in detection of unknown types of attacks in microphone channel. However, similar task in telephone channel remains extremely relevant, for example, in the banking sector. Method. The aim of the work was to study the applicability of deep learning approach for described problem solution, in particular, convolutional neural networks with the Max-Feature-Map activation function.Main Results.The experiments performed for real telephone attacks showed insufficient efficiency of the systems trained on data with emulated telephone channel. That is why, the database of real spoofing attacks in telephone channel was collected. The best system demonstrated 1.5% equal error rate (EER) on a subset of replay attacks, 1.7% for voice conversion attacks, and 2.8% for attacks with voice synthesis. Experiments show the need to consider different recording conditions, due to the great number of factors that have the influence on the channel. Practical Relevance.The results of the work can be applied in the field of voice biometrics. The presented methods can be used in systems of automatic speaker verification and identification for detection of spoofing attacks on them.

The concept of globalization or internationalization of certain wars, which were result of terrorist activities worldwide, as well as the high attention of terrorism coverage worldwide broadcasting might open up better opportunities to journalists – particularly to those who work in democratic countries like U.S.A and India – to improve their coverage. The context is the key: the context of the operation methodology, follow of regulatory bodies guidelines, the journalistic culture and the global environment. It is very important how media presents consequences of terrorist acts, how information is transmitted to public. Television and press have had a significant impact on how public receives terrorist acts and their consequences. As a result, nearly each public survey indicates that responders almost anywhere put fear of terrorist acts on the top of their priority list. In order to reach out a conclusion on this paper, a researcher has gone through number of books related to terrorism and media, has examined significant number of journals which deals with core issue of terrorism and its coverage by media.