Research Interests:
The task of pattern recognition is a task of division of a feature space into regions separating the training examples belonging to different classes. Support Vector Machines (SVM) identify the most borderline examples called support... more
The task of pattern recognition is a task of division of a feature space into regions separating the training examples belonging to different classes. Support Vector Machines (SVM) identify the most borderline examples called support vectors and use them to determine discrimination hyperplanes (hyper–curves). In this paper a pattern recognition method is proposed which represents an alternative to SVM algorithm. Support vectors are identified using selected methods of computational geometry in the original space of features i.e. not in the transformed space determined partially by the kernel function of SVM. The proposed algorithm enables usage of kernel functions. The separation task is reduced to a search for an optimal separating hyperplane or a Winner Takes All (WTA) principle is applied.
Research Interests:
Lung ultrasound is used to detect various artifacts in the lungs that support the diagnosis of different conditions. There is ongoing research to support the automatic detection of such artifacts using machine learning. We propose a... more
Lung ultrasound is used to detect various artifacts in the lungs that support the diagnosis of different conditions. There is ongoing research to support the automatic detection of such artifacts using machine learning. We propose a solution that uses analytical computer vision methods to detect two types of lung artifacts, namely A- and B-lines. We evaluate the proposed approach on the POCUS dataset and data acquired from a hospital. We show that by using the Fourier transform, we can analyze lung ultrasound images in real-time and classify videos with an accuracy above 70%. We also evaluate the method’s applicability for segmentation, showcasing its high success rate for B-lines (89% accuracy) and its shortcomings for A-line detection. We then propose a hybrid solution that uses a combination of neural networks and analytical methods to increase accuracy in horizontal line detection, emphasizing the pleura.
Research Interests:
Research Interests:
Research Interests:
Research Interests:
ABSTRACT The paper deals with a comparison study of support vector machines classification approach and ARTMAP neural classifiers. SVM provides very interesting mathematical methods based on virtual transformation of input space into a... more
ABSTRACT The paper deals with a comparison study of support vector machines classification approach and ARTMAP neural classifiers. SVM provides very interesting mathematical methods based on virtual transformation of input space into a multidimensional space. The high degree of nonlinear discrimination hyperplane is approximated by task transformation into dichotomical classification with the aim to achieve the best classification results. SVM was used with RBF kernel and experiments were done on benchmark data as well as on real-world datellite images over Slovakia. Comparisons with Fuzzy Artmap and Gaussian Artmap on these data were accomplished. Adaptive kernel function based on neural network is proposed for future research in this area. Classification is evaluated using contigency tables for multiclass classification problems. The aim was to develop a classification tool with the highest accuracy on the tested images.
Research Interests:
Research Interests:
This article focuses on simple rehabilitation video-game called Flying with Friends. The rehabilitation industry has been experiencing a boom in recent years, coupled with the growing popularity of virtual reality technology, a drop in... more
This article focuses on simple rehabilitation video-game called Flying with Friends. The rehabilitation industry has been experiencing a boom in recent years, coupled with the growing popularity of virtual reality technology, a drop in prices for these technologies and the expansion entertainment industry in the form of computer games. The goal of this experiment was to provide a proof that such systems are viable option when it comes to artificial intelligence systems in serious video-games, but not limited to only serious ones. The solution described in this article, in cooperation with experts, is going to be deployed in a real rehabilitation environment.
Research Interests:
Research Interests:
Wrist is one of the most complicated parts of the human's musculoskeletal system. It is not made up of one, but of a series of small joints that bind the bones of the forearm with the palm bone. Because of its complex morphology, and... more
Wrist is one of the most complicated parts of the human's musculoskeletal system. It is not made up of one, but of a series of small joints that bind the bones of the forearm with the palm bone. Because of its complex morphology, and its frequent use in almost every activity, wrist is extremely prone to damage. Small wrist injuries can completely change one's life, and sometimes they cannot even be avoided. These include, for example inflammatory wrist syndrome, recurrent use syndrome that occurs after multiple repetitions of the same activity as well as the carpal tunnel syndrome. Most of these minor damages can be healed by wrappings, medications, or reductions, or by complete removal of the load on the wrist for a certain period. Rehabilitation of such injuries is simple, so it is often home remedy. Problems are the injuries that cause serious damage to bones wrist binder. These often require fixation, in many cases connected with surgical procedure and direct fixation of bones by internal or external stabilizers. After these steps, which are used for restoring the moving mobility of the wrist, computer assisted rehabilitation can be useful. In this article, we want to describe modern trends in computer assisted wrist rehabilitation.
Research Interests:
Certain post-thoracic surgery complications are monitored in a standard manner using methods that employ ionising radiation. A need to automatise the diagnostic procedure has now arisen following the clinical trial of a novel lung... more
Certain post-thoracic surgery complications are monitored in a standard manner using methods that employ ionising radiation. A need to automatise the diagnostic procedure has now arisen following the clinical trial of a novel lung ultrasound examination procedure that can replace X-rays. Deep learning was used as a powerful tool for lung ultrasound analysis. We present a novel deep-learning method, automated M-mode classification, to detect the absence of lung sliding motion in lung ultrasound. Automated M-mode classification leverages semantic segmentation to select 2D slices across the temporal dimension of the video recording. These 2D slices are the input for a convolutional neural network, and the output of the neural network indicates the presence or absence of lung sliding in the given time slot. We aggregate the partial predictions over the entire video recording to determine whether the subject has developed post-surgery complications. With a 64-frame version of this archit...
Research Interests:
Research Interests:
We have build a device named RepaiR for strength measurements and isokinetic rehabilitation of the wrist joint. We have performed series of measurements on 25 healthy individuals and 10 patients with neuromuscular and traumatic... more
We have build a device named RepaiR for strength measurements and isokinetic rehabilitation of the wrist joint. We have performed series of measurements on 25 healthy individuals and 10 patients with neuromuscular and traumatic impairments. Our initial goal was to verify that the measured data contain sufficient information to distinguish between healthy and not healthy subjects as a proof of concept. We have implemented Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM) that processes the time structured measurements. LSTM effectively models the varying length of the input vector and the long term dependencies. We compare performances of our models on the data sets with varying minimal input vector lengths. We have proven that the measurements can be used to detect neuromuscular impairments and our best performing model worked with 77,6 % accuracy.
Research Interests:
In this article, we present our current application research regarding measurement and processing of Electromyography data subsequently used for gesture detection. The rehabilitation area has been experiencing a huge progress in recent... more
In this article, we present our current application research regarding measurement and processing of Electromyography data subsequently used for gesture detection. The rehabilitation area has been experiencing a huge progress in recent years. This is due to an increase in the number of patients with various types of disability, technological advance and large number of available devices. One of the rehabilitation sub-areas is the rehabilitation of patients with motor impairment.This type of rehabilitation often involves different virtual reality or augmented reality systems. Such systems are in need of accurate, inexpensive and user-friendly devices acting as controllers. In this paper we propose a system controlled via electromyography. This system is designed to aid rehabilitation of patients with impaired finger control and movements. The electromyography data are measured by Myo bracelet, processed and classified by support vector machines classifier during the offline training.
Research Interests:
This article focuses on the issue of rehabilitation of patients with motor impairment using computer games. The rehabilitation industry has been experiencing a boom in recent years, coupled with the growing popularity of virtual reality... more
This article focuses on the issue of rehabilitation of patients with motor impairment using computer games. The rehabilitation industry has been experiencing a boom in recent years, coupled with the growing popularity of virtual reality technology, a drop in prices for these technologies and the expansion the entertainment industry in the form of computer games. We are trying to create a computer game system designed for home rehabilitation of patients. The aim of this article is to devise with a subset of motor impairment assessment of patients with the intention of adapting the demands of computer games based on the physical abilities of a particular patient. We are all trying to keep this system at the level available to a regular home user.
Research Interests:
This study deals with the problems of aggregating the opinions of a group of people in such a way that the quality of the group decision surpasses the quality of the decision of the most experienced individual within the group. The... more
This study deals with the problems of aggregating the opinions of a group of people in such a way that the quality of the group decision surpasses the quality of the decision of the most experienced individual within the group. The methods we have studied fall in the research domain of the so called collective intelligence. We provide an overview of the state-of-the-art in the collective intelligence. We describe the method based on adaptive boosting we have proposed aggregatig the opinions of a group of people. We have implemented a web application to gather opinions of people and used the application to collect data for the experimental analysis. The model problem was to identify whether there is or there is not a tumor present in the series of X-ray images of human lungs. We have compared our proposed method to conventional methods such as majority voting. We have concluded that our proposed method can be successfully used to aggregate opinions of a group of people to increase th...
Research Interests:
Hand and wrist tremors are rhythmic movement with oscillations of one or more body parts. Most common is the Parkinsonian tremor, that affects approximately 3.8 million people around the world. In Slovakia, 750 to 1,000 patients with... more
Hand and wrist tremors are rhythmic movement with oscillations of one or more body parts. Most common is the Parkinsonian tremor, that affects approximately 3.8 million people around the world. In Slovakia, 750 to 1,000 patients with Parkinson's disease are diagnosed annually. The purpose of this article is to describe the measurement of tremors using force measurement sensors. With frequent diagnostics, we can detect essential tremor (ET) in patients and apply appropriate treatment.
Research Interests:
The underlying idea of the research presented here is connecting the intelligent entities — either living or artificial — into systems that act as a unit and that are smarter than their smartest individual elements. Collective decision... more
The underlying idea of the research presented here is connecting the intelligent entities — either living or artificial — into systems that act as a unit and that are smarter than their smartest individual elements. Collective decision making is common in human society. Human is a social being that conforms to the group or society. Decisions taken collectively may be better or worse than decisions taken by a selected individual and we search for methods of making better collective decisions. Our world is abundant with opinions and some of these were even produced by machines. We focus here on cooperation of intelligent entities in hybrid collectives. The application domains are aggregation of opinions using methods of computational intelligence and adaptive computer games in rehabilitation of motoric impairments. The reader is introduced to the state of the art in the fields of collective intelligence and computer aided rehabilitation. We present our research in the relevant fields ...
Research Interests:
A new approach of the string theory called the Prediction Model Based on String Invariants (PMBSI) was applied here to time-series forecast. We used 2-end-point open string that satisfies the Dirichlet and Neumann boundary conditions. The... more
A new approach of the string theory called the Prediction Model Based on String Invariants (PMBSI) was applied here to time-series forecast. We used 2-end-point open string that satisfies the Dirichlet and Neumann boundary conditions. The initial motivation was to transfer modern physical ideas into the neighboring field called econophysics. The physical statistical viewpoint has proved to be fruitful, namely in the description of systems where many-body effects dominate. However, PMBSI is not limited to financial forecast. The main advantage of PMBSI include absence of the learning phase when large number of parameters must be set. Comparative experimental analysis of PMBSI vs. SVM was performed and the results on artificial and real-world data are presented. PMBSI performance was in a close match with SVM.
Research Interests:
Airborne LiDAR produced large amounts of data for archaeological research over the past decade. Labeling this type of archaeological data is a tedious process. We used a data set from Pacunam LiDAR Initiative survey of lowland Maya region... more
Airborne LiDAR produced large amounts of data for archaeological research over the past decade. Labeling this type of archaeological data is a tedious process. We used a data set from Pacunam LiDAR Initiative survey of lowland Maya region in Guatemala. The data set contains ancient Maya structures that were manually labeled, and ground verified to a large extent. We have built and compared two deep learning-based models, U-Net and Mask R-CNN, for semantic segmentation. The segmentation models were used in two tasks: identification of areas of ancient construction activity, and identification of the remnants of ancient Maya buildings. The U-Net based model performed better in both tasks and was capable of correctly identifying 60–66% of all objects, and 74–81% of medium sized objects. The quality of the resulting prediction was evaluated using a variety of quantifiers. Furthermore, we discuss the problems of re-purposing the archaeological style labeling for production of valid machi...
Research Interests: Computer Science, Artificial Intelligence, Remote Sensing, Mesoamerican Archaeology, LiDAR, and 12 moreMaya Archaeology, Remote sensing and GIS applications in Landscape Research, Mayan Studies, Artificial Neural Networks for modeling purposes, Maya, Artificial Neural Networks, Segmentation, LiDAR for Landscape Archaeology, Inteligencia artificial, Classic Maya archaeology, Point Cloud, and LiDAR for Archaeology
An original swarm-based method for coordination of groups of mobile robots with a focus on the self-organization and self-adaptation of the groups is presented in this paper. The method is a nature-inspired decentralized algorithm that... more
An original swarm-based method for coordination of groups of mobile robots with a focus on the self-organization and self-adaptation of the groups is presented in this paper. The method is a nature-inspired decentralized algorithm that uses artificial pheromone marks and enables the cooperation of different types of independent reactive agents that operate in the air, on the ground, or in the water. The advantages of our solution include scalability, adaptability, and robustness. The algorithm worked with variable numbers of agents in the groups. It was resistant against failures of the individual robots. A transportation control algorithm that ensured the spreading of different types of agents across exploration space with different types of environments was introduced and tested. We established that our swarm control algorithm was able to successfully control three basic behaviors: space exploration, population management, and transportation. The behaviors were able to run simulta...
Research Interests:
We have developed a device, the Rehapiano, for the fast and quantitative assessment of action tremor. It uses strain gauges to measure force exerted by individual fingers. This article verifies the device’s capability to measure and... more
We have developed a device, the Rehapiano, for the fast and quantitative assessment of action tremor. It uses strain gauges to measure force exerted by individual fingers. This article verifies the device’s capability to measure and monitor the development of upper limb tremor. The Rehapiano uses a precision, 24-bit, analog-to-digital converter and an Arduino microcomputer to transfer raw data via a USB interface to a computer for processing, database storage, and evaluation. First, our experiments validated the device by measuring simulated tremors with known frequencies. Second, we created a measurement protocol, which we used to measure and compare healthy patients and patients with Parkinson’s disease. Finally, we evaluated the repeatability of a quantitative assessment. We verified our hypothesis that the Rehapiano is able to detect force changes, and our experimental results confirmed that our system is capable of measuring action tremor. The Rehapiano is also sensitive enough...
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Turntables continue to be manufactured and sold today, although in small numbers. Some audiophiles still prefer the sound of vinyl records over that of digital music sources and are ready to pay for high end gramophone systems. We... more
Turntables continue to be manufactured and sold today, although in small numbers. Some audiophiles still prefer the sound of vinyl records over that of digital music sources and are ready to pay for high end gramophone systems. We pre-sent here a tonearm constructed so that its rotating cartridge is always nearly tan-gent to the record's groove the stylus is reading thus greatly reducing the anti-skating force. The physical dimensions of the tonearm are optimized using genetic algorithm to minimize the deviation of the cartridge from the tangent along its entire trajectory (tangent tracking). We describe the specific optimization algorithm used, report the simulated results and present a working prototype.
Research Interests:
Research Interests:
Selected computational geometry algorithms based on De-launay graph (triangulation) and Support Vector Machines used in pat-tern recognition applications divide feature space into regions surround-ing patterns representing separate... more
Selected computational geometry algorithms based on De-launay graph (triangulation) and Support Vector Machines used in pat-tern recognition applications divide feature space into regions surround-ing patterns representing separate classes. The approaches to the feature space division are analogous, the results are usually similar and the clas-sification processes differ. The selected algorithms were variously pre-set and analyzed in order to evaluate the measure of similarity among them. The purpose was to identify possible ways of hybridization while a syn-ergic effect was to be achieved. Properties of the individual algorithms were compared using several cri-teria. Artificial and real world data were used. Differences regarding the mechanism of selection and location of the support vectors were de-scribed. Optimality of the resulting decision boundaries was assessed. Robustness and accuracy of the algorithms were compared as well. The computational geometry based algorithms were ...
ABSTRACT A gradient based learning for ANN training in pattern recognition tasks and a genetic approach for ANN pruning are proposed in this paper. The goal is to achieve a wide margin classifier the Vapnik-Chevornenkis (VC) dimension of... more
ABSTRACT A gradient based learning for ANN training in pattern recognition tasks and a genetic approach for ANN pruning are proposed in this paper. The goal is to achieve a wide margin classifier the Vapnik-Chevornenkis (VC) dimension of which is being reduced in order to increase the generalization performance. Inspired by Support Vector Machines the examples closest to the decision boundary contribute to the training the most. The training penalty is rule-based and calculated according to the spatial distribution of the training examples relative to the separating hyperplane. The tendency to saturation of hidden neurons is suppressed. Genetic algorithm based method is proposed for reduction of the size of a trained ANN. The proposed algorithms were tested on artificial and real world data and compared to standard Backpropagation and Support Vector Machine with Gaussian RBF kernel.
Research Interests:
ABSTRACT Growing ensemble of linear classifiers uses the 'divide and conquer' strategy in pattern recognition tasks. Performing the competitive learning the feature space is divided into subregions where linear classifiers... more
ABSTRACT Growing ensemble of linear classifiers uses the 'divide and conquer' strategy in pattern recognition tasks. Performing the competitive learning the feature space is divided into subregions where linear classifiers are constructed. The structure of the ensemble is growing during the training and it is self determined. The overall output of the ensemble is the output of a winning member. The method is not bound to a specific type of classifier. According to the experimental results achieved on artificial and real world datasets the algorithm performs comparably to Gaussian SVM. Due to the simple nature of the decision boundary, knowledge retrieval procedures can be applied.
Research Interests:
We use Tracking-Learning-Detection algorithm (TLD) [1]-[3] to localize and track objects in images sensed simultaneously by two parallel cameras in order to determine 3D coordinates of the tracked object. TLD method was chosen for its... more
We use Tracking-Learning-Detection algorithm (TLD) [1]-[3] to localize and track objects in images sensed simultaneously by two parallel cameras in order to determine 3D coordinates of the tracked object. TLD method was chosen for its state-of-art performance and high robustness. TLD stores the object to be tracked as a set of 2D grayscale images that is incrementally built. We have implemented the 3D tracking system into a PC, communicating with the Nao humanoid robot [4][5] equipped with a stereo camera head. Experiments evaluating the accuracy of the 3D tracking system are presented. The robot uses feed-forward control to touch the tracked object. The controller is an artificial neural network trained using the error Back-Propagation algorithm. Experiments evaluating the success rate of the robot touching the object are presented.
Research Interests:
ABSTRACT In this paper we present the feed-forward neural network controller of robotic arm, which makes use of tracking method applied to stereo-vision cameras mounted on the head of the humanoid robot Nao, in order to touch the tracked... more
ABSTRACT In this paper we present the feed-forward neural network controller of robotic arm, which makes use of tracking method applied to stereo-vision cameras mounted on the head of the humanoid robot Nao, in order to touch the tracked object. The Tracking-Learning-Detection (TLD) method, which we use to detect and track the object, is known for its state-of-art performance and high robustness. This method was adjusted to be usable with a stereo-vision camera system, in order to provide 3D spatial coordinates of the object. These coordinates are used as the input for the feed-forward controller, which controls the arm of a humanoid robot. The goal of the controller is to move the hand of the robot to the object by setting arm joints into position corresponding to the object location. The controller is implemented as an artificial neural network and trained using the error back-propagation algorithm. The experiment, which demonstrates the proof of the concept, is also denoted in this paper.
Research Interests:
Research Interests:
ABSTRACT This paper describes a method for production of an ensemble of general classifiers using unsupervised learning. The method uses the ‘divide and conquer’ strategy. Using competitive learning the feature space is divided into... more
ABSTRACT This paper describes a method for production of an ensemble of general classifiers using unsupervised learning. The method uses the ‘divide and conquer’ strategy. Using competitive learning the feature space is divided into subregions where the classifiers are constructed. The structure of the ensemble starts from a single member and new members are being added during the training. The growth of the ensemble is self determined until the ensemble reaches the desired accuracy. The overall response of the ensemble to an input pattern is represented by the output of a winning member for the particular pattern. The method is generic, i.e. it is not bound to a specific type of classifier and it is suitable for parallel implementation. It is possible to use the method for data mining. A basic wide margin linear classifier is used in the experiments here. Experimental results achieved on artificial and real world data are presented and compared to the results of Gaussian SVM. Parallel implementation of the method is described.
Research Interests:
This paper deals about sensor network and its communication and energy consumption reduction. Our implementation is focused on reaching communication reduction to spare power resources on sensor nodes. To decrease amount of communication... more
This paper deals about sensor network and its communication and energy consumption reduction. Our implementation is focused on reaching communication reduction to spare power resources on sensor nodes. To decrease amount of communication and not decrease amount information about monitored or controlled system we used data mining technique and events based communication. Data mining technique was implemented directly on sensor node, which offers enough storage and computing sources. Implemented data mining technique is SPRT (Sequential Probability Ratio Test). Using this method, the sensor decides, if the monitored system is behaving abruptly. Like the result of decision, the sensor creates appropriate type of event. The sensors are subscribed to event listener, which handle these events. Using events in sensor network, with transmitting only valuable information, has good result in communication reduction and faster feedback from monitored system and with information.