Search

In this paper, we evaluated the advantages of using the Hilbert transform in an optical correlator. We reviewed state-of-art optical techniques that employ the Hilbert transform and, based on the principle of the classic Hilbert correlator, we performed simulations for this type of optical correlator. The simulation results confirm the efficiency of the Hilbert transform optical correlator for image recognition applications and certify the suitability of using this type of optical correlator in future applications, as a viable alternative to other techniques, either optical or digital.

Artificial Neural Networks (ANN) are modelled on the workings of the human brain where information is transported through the synapses formed between individual neurons. The activation function decides the signal to be transmitted from a node of the ANN to the others that are connected, so we can interpret the value of the function as the "activation" of a neuron. There are many activation functions available, each with its advantages and disadvantages when it comes to learning rate and computational load. The current article presents the most commonly used activation functions and their applications in ANNs. We study the effect of using different activation functions for a convolutional neural network used for image classification. Model performance for the train and test set are presented for each fold of the k-fold cross-validation, as well as accuracy scores for the trained network.

Estimating precise metric depth is an essential task for UAV navigation, which is very difficult to learn unsupervised without access to odometry. At the same time, depth recovery from kinematics and optical flow is mathematically precise, but less numerically stable and robust, especially in the focus of expansion areas. We propose a model that combines the analytical, vision-with-odometry approach, with deep unsupervised learning, into a single formulation for metric depth estimation, which is both fast and accurate. The two pathways – analytical and data-driven – form a robust ensemble, which provides supervisory signal to a single deep net that distills the consensus between scene geometry, pose, kinematics, camera intrinsics and the input RGB. The distilled net has low runtime and memory costs, being suitable for embedded devices. We validate our results against an off-the-shelf SfM-based solution. We also introduce a new real-world dataset of almost 20 minutes of continuous UAV flight, on which we demonstrate better accuracy and capabilities than the deep learning and analytical approaches.

Several important issues are affecting the forest environment due to deforestation and natural disasters (for example forest fires, or increased gas emissions). This paper proposes an intelligent forest environment monitoring solution based on the Raspberry Pi Model 3, analogical and digital sensors and signals analysis algorithms. Parameters such as temperature, gas concentrations, soil humidity etc. are monitored with sensors while background sounds are analyzed with a classification algorithm on the basis of which the generated event can be classified into one of the following categories: Chainsaw, Vehicle, or Forest background noise. The user’s accessibility to the collected data is ensured via Internet and a mobile applications that allows the user to receive notifications, whenever fire, pollution sources, or illegal deforestation are detected. The SeaForest environment monitoring solution is an IoT project, addressed to public and private forest owners as well as to national environmental and disaster response authorities.

Automotive industry is a very large industry across the globe and nowadays all about designing and manufacturing innovative solutions are for a safe, comfortable, and efficient driving experience. It is important to include radar sensors in driver assistance systems because can increase the safety for the driver, passengers, and pedestrians. The paper is focused on communication between radar sensors, presents a new approach for transmitting large data between the nodes. In order to test if the radar sensors are well packed and prepared for commercial use, a software solution is offered.

Waking up in the morning represents a daily routine activity for the most part of the people who use common sound alarm clocks or who use their mobile phone as an alarm clock. For hearing-impaired people these solutions are not suitable. Other systems are equipped also with vibration motors and, thus, let the user choose if vibrations or sound signal are triggered when the alarm starts. These systems, though, require the time adjustment whenever the device is unplugged or when power outage takes place. Moreover, there is no solution available that gives a feedback to the caregivers, in order to be noticed that the user has been indeed waken by the alarm system. Through this paper, we propose an alarm clock system based on vibrations motors, ultrasonic sensors, Bluetooth and LoRa technology and server to wake the user, to ascertain the awakening and to send notifications to the caregivers. In addition, to mitigate the need to adjust the time in the case of a power failure, through the connection to LoRa server and to the battery operation, the time can be automatically updated.

As technology is evolving day by day, the remote control is starting to be a normality in our lives. In this context, it is described an implementation of a system for controlling the functional time of consumers in a living apartment for daily use. It consists in an electronic module and a soon to be mobile application running on Android devices connected through Bluetooth. The functionality of the system consists in allowing the user to program the running time for a consumer in his home by entering the times in the application.

This paper deals with intraday/close to real time multilateral regional redispatching optimization, with the special focus on the effects of the inclusion of distributed electricity prosumers in redispatching actions. The redispatching concept includes the co-optimization of technical (sensitivity factors based on network locations) and economic criteria (redispatch bid price) for the selection and application of optimal redispatching actions to solve the network congestions, while seeking for the potential of flexible distributed electricity prosumers to participate in redispatching mechanism. Based on the drafted concepts for cross-border redispatching optimisation, and by using the developed software tools, regional redispatching analysis is performed on a small test case network model with observed overloads.

LEDs tend to become the conventional lighting solutions. Their property of being able to be quickly switched on and off inspired scientists to develop new communication systems, like Visible Light Communication (VLC), in which the transmitter is usually represented by LEDs and the receiver by photodetectors. Optical Camera Communications (OCC) is similar to VLC, but the receiver is an image sensor. A great challenge in OCC is to design a communication system that can use low frame rate cameras while avoiding the occurrence of light flickering at the transmitter. This paper proposes and experimentally validates an OCC solution that uses a camera capturing 30 frames per second as receiver, does not produce flickering at the transmitter, and, moreover, does not require a transmitter-receiver synchronization.

Healthcare marketing is one of the most interesting research domains due to the high complexity and legal regulations of the medical data on the one side and the large volume of data required by marketing processes on the other side. Current research proposes a solution of adapting a commonly used marketing process, recommendation to healthcare industry needs to healthcare industry. Optimization of the standard recommendation processes have been identified as the main point of adaptation due to the high volume of data available in healthcare systems along with the need of analyzing the complete set of available data. Caching results represents one of the most common optimization techniques used in software implementation and due to highly sensitivity of medical data appears to be non-applicable. This paper is making use of segmentation of data sets combining patient profiles in target groups that are to be passed to recommendation scenarios as input data allowing catching results for healthcare recommendation scenarios.

The education system is a source that generates significant amounts of data, daily, in various formats and, often, hiding valuable information. Finding a good way to unravel those hidden gems, represents one of the most challenging problems of natural language processing, namely sentiment analysis. This involves, applying NLP and text analysis techniques to identify and classify subjective opinions in different materials such as documents or sentences. In our work, we used as raw data, the opinions of students from eleven high schools in Suceava, related to various aspects of the educational process. They were collected through a Google Docs form, and analyzed through the Orange environment (an open source tool for machine learning and data visualization). In this paper, we make a comparative study of the obtained results using the Ekman and Plutchik models. Each model extracts from the analyzed texts, a different emotion, based on which the students’ sentiments towards the educational process will be analyzed.

Grid Locator systems describe locations anywhere in the world and are mainly used in radio communication to transmit stations location of the correspondents. This paper proposes to extend the use of Maidenhead Locator System to mobile robots navigation by transforming the GPS coordinates into higher precision extended grid squares. The experimental part of the paper was validated using real GPS data from a commercial drone. The data, as well as the source for the reproduction of the experimental part, is available at the homepage of the authors.