Search

The paper presents the analysis of modern Artificial Intelligence algorithms for the automated system supporting human beings during their conversation in Polish language. Their task is to perform Automatic Speech Recognition (ASR) and process it further, for instance fill the computer-based form or perform the Natural Language Processing (NLP) to assign the conversation to one of predefined categories. The state-of-the-art review is required to select the optimal set of tools to process speech in the difficult conditions, which degrade accuracy of ASR. The paper presents the top-level architecture of the system applicable for the task. Characteristics of Polish language are discussed. Next, existing ASR solutions and architectures with the End-To-End (E2E) Deep Neural Network (DNN) based ASR models are presented in detail. Differences between Recurrent Neural Networks (RNN), Convolutional Neural Networks (CNN) and transformers in the context of ASR technology are also discussed. [ABSTRACT FROM AUTHOR]

The aim of the paper is to present the distributed system for the unwanted event detection regarding inmates in the closed penitentiary facilities. The system processes large number of data streams from IP cameras (up to 180) and performs the event detection using Deep Learning neural networks. Both audio and video streams are processed to produce the classification outcome. The application-specific data set has been prepared for training the neural models. For the particular event types 3DCNN and YOLO architectures have been used. The system was thoroughly tested both in the laboratory conditions and in the actual facility. Accuracy of the particular event detection is on the satisfactory level, though problems with the particular events have been reported and will be dealt with in the future. [ABSTRACT FROM AUTHOR]

The paper presents the state of the art and advancement of artificial intelligence methods in analog systems diagnostics. Firstly, the diagnostic domain is introduced and its problems explained. Then, computational intelligence approaches usable for fault detection and identification are reviewed. Particular groups of methods are presented in detail, explaining their usefulness and drawbacks. Examples, such as the induction motor or the electronic filter, are provided to show the applicability of the presented approaches for monitoring the state of analog objects from engineering domains. The discussion section reviews the presented approaches, their future prospects and problems to be solved.

In this paper, the secure low-power Internet of Things (IoT) transmission methods for encryption and digital signature are presented. The main goal was to develop energyefficient method to provide IoT devices with data confidentiality, integrity, and authenticity. The cryptograph energy efficient and security algorithms modifications for IoT domain were made. The novelty in our solution is the usage of encryption method popular in the image processing in the domain of the Internet of Things. Proposed modification improves immunity for the brute-force and plain-text attacks. Furthermore, we propose the modifications for hash calculation method to transform it into digital signature calculation method that is very sensitive to input parameters. The results indicate low energy consumption of both methods, however it varies significantly depending on the architecture of the devices. [ABSTRACT FROM AUTHOR]

The paper presents the analysis of the Commercial Off-The-Shelf (COTS) software regarding the ability to be used in audio steganography techniques. Such methods are a relatively new tool for hiding and transmitting crucial information, also being used by hackers. In the following work, the publicly available software dedicated to audio steganography is examined. The aim was to provide the general operating model of the information processing in the steganographic effort. The embedding method was analyzed for each application, providing interesting insights and allowing classifying the methods. The results prove that it is possible to detect the hidden message within the specific audio file and identify the technique that was used to create it. This may be exploited further during the hacking attack detection and prevention. [ABSTRACT FROM AUTHOR]

Series arc faults cause the majority of household fires involving electrical failures or malfunctions. Low-fault current amplitude is the reason for the difficulties faced in implementing effective arc detection systems. The paper presents a novel arc detection and faulty line identification method. It can be easily used in the low-voltage Alternate Current (AC) household network for arc detection in the Non-Intrusive Load Monitoring (NILM). Unlike existing methods, the proposed approach exploits both current and voltage signal time domain analysis. Experiments have been conducted with up to six devices operating simultaneously in the same circuit with an arc fault generator based on the IEC 62606:2013 standard. Sixteen time-domain features were used to maximize the arc-fault detection accuracy for particular appliances. Performance of the random forest classifier for arc fault detection was evaluated for 28 sets of features with five different sampling rates. For the single period analysis arc, detection accuracy was 98.38%, with F-score of 0.9870, while in terms of the IEC 62606:2013 standard, it was 99.07%, with F-score of 0.9925. Location of a series arc fault (line selection) was realized by identifying devices powered by the faulty line. The line selection was based on the Mean Values of Changes feature vector (MVC50), calculated for absolute values of differences between adjacent current signal periods during the arc fault. The fault location accuracy was 93.20% for all cases and 98.20% for cases where the arc fault affected a single device. [ABSTRACT FROM AUTHOR]

The paper presents the novel concept of the magnetoelectric sensor constructed using the amorphous glass ribbon. Its output characteristics (voltage pattern), conditions of work and experimental results are presented. The novel construction allows for minimizing the demagnetizing field in the core of the sensor and linearization of the characteristics between the magnetic field and obtained voltage. Conducted experiments were aimed at determining the sensor operation in the presence of the constant magnetic field (HDC). The main concern of the tests was to verify the linear dependency between the HDC value and the amplitude of the output voltage. Next, the computer model representing the sensor behavior in the constant magnetic field is described. The model implements the parameter identification task based on the regression algorithms. The presented work shows that the proposed device can be used to measure the weak magnetic field and the dependency between the output signal amplitudes and the constant component in the measured magnetic field is approximately linear. This enables measurements of even weak fields. [ABSTRACT FROM AUTHOR]

Copyright of Przegląd Elektrotechniczny is the property of Przeglad Elektrotechniczny and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The paper presents a novel heuristic procedure (further called the AH Method) to investigate function shape in the direct vicinity of the found optimum solution. The survey is conducted using only the space sampling collected during the optimization process with an evolutionary algorithm. For this purpose the finite model of point-set is considered. The statistical analysis of the sampling quality based upon the coverage of the points in question over the entire attraction region is exploited. The tolerance boundaries of the parameters are determined for the user-specified increase of the objective function value above the found minimum. The presented test-case data prove that the proposed approach is comparable to other optimum neighborhood examination algorithms. Also, the AH Method requires noticeably shorter computational time than its counterparts. This is achieved by a repeated, second use of points from optimization without additional objective function calls, as well as significant repository size reduction during preprocessing. [ABSTRACT FROM AUTHOR]

Copyright of Farmacia is the property of Societatea de Stiinte Farmaceutice Romania and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)