Search

Background: Motor imagery is a cognitive process of imagining a performance of a motor task without employing the actual movement of muscles. It is often used in rehabilitation and utilized in assistive technologies to control a brain–computer interface (BCI). This paper provides a comparison of different time–frequency representations (TFR) and their Rényi and Shannon entropies for sensorimotor rhythm (SMR) based motor imagery control signals in electroencephalographic (EEG) data. The motor imagery task was guided by visual guidance, visual and vibrotactile (somatosensory) guidance or visual cue only. Results: When using TFR-based entropy features as an input for classification of different interaction intentions, higher accuracies were achieved (up to 99.87%) in comparison to regular time-series amplitude features (for which accuracy was up to 85.91%), which is an increase when compared to existing methods. In particular, the highest accuracy was achieved for the classification of the motor imagery versus the baseline (rest state) when using Shannon entropy with Reassigned Pseudo Wigner–Ville time–frequency representation. Conclusions: Our findings suggest that the quantity of useful classifiable motor imagery information (entropy output) changes during the period of motor imagery in comparison to baseline period; as a result, there is an increase in the accuracy and F1 score of classification when using entropy features in comparison to the accuracy and the F1 of classification when using amplitude features, hence, it is manifested as an improvement of the ability to detect motor imagery. [ABSTRACT FROM AUTHOR]

In this paper, we present automatic, deep-learning methods for pipeline detection in underwater environments. Seafloor pipelines are critical infrastructure for oil and gas transport. The inspection of those pipelines is required to verify their integrity and determine the need for maintenance. Underwater conditions present a harsh environment that is challenging for image recognition due to light refraction and absorption, poor visibility, scattering, and attenuation, often causing poor image quality. Modern machine-learning object detectors utilize Convolutional Neural Network (CNN), requiring a training dataset of sufficient quality. In the paper, six different deep-learning CNN detectors for underwater object detection were trained and tested: five are based on the You Only Look Once (YOLO) architectures (YOLOv4, YOLOv4-Tiny, CSP-YOLOv4, YOLOv4@Resnet, YOLOv4@Den seNet), and one on the Faster Region-based CNN (RCNN) architecture. The models' performances were evaluated in terms of detection accuracy, mean average precision (mAP), and processing speed measured with the Frames Per Second (FPS) on a custom dataset containing underwater pipeline images. In the study, the YOLOv4 outperformed other models for underwater pipeline object detection resulting in an mAP of 94.21% with the ability to detect objects in real-time. Based on the literature review, this is one of the pioneering works in this field. [ABSTRACT FROM AUTHOR]

This paper approaches the problem of signal denoising in time-variable noise conditions. Non-stationary noise results in variable degradation of the signal's useful information content over time. In order to maximize the correct recovery of the useful part of the signal, this paper proposes a denoising method that uses a criterion based on amplitude segmentation and local Rényi entropy estimation which are limited over short time blocks of the signal spectrogram. Local estimation of the signal features reduces the denoising problem to the stationary noise case. Results, presented for synthetic and real data, show consistently better performance gained by the proposed adaptive method compared to denoising driven by global criteria. [ABSTRACT FROM AUTHOR]

Understanding the dynamics of tidal rivers and estuaries is critical for reliable water management. Recently, the use of Machine Learning (ML) has increased in favor of hydrologic and hydraulic models. The advantages of ML over physically based models are most evident in modeling complex and nonlinear hydrologic processes and inverse problems. This study provides a critical review of ML approaches for forecasting, reconstruction, and establishment of stage-discharge relationships in tidal rivers and estuaries characterized by nonlinear interaction between the river and coastal processes. Gaps in this research area and the limited number of stage-discharge studies are identified and explained. The advantages and limitations of each approach are discussed from a critical perspective, and suggestions are made for future research directions. Advanced Recurrent Neural Networks (RNNs) and hybrid modeling systems combining physically based models with ML appear to be the most promising approaches for modeling complex physical processes in these environments. • A comprehensive review of machine learning techniques for predicting water level and discharge. • Focused on tidal rivers and estuaries characterized by nonlinear interactions. • Discussion on limitations and advantages of the machine learning approaches. • Presentation of hybrid approaches that performed better than just machine learning models. • Recommendations for future research directions. [ABSTRACT FROM AUTHOR]

This paper proposes an automatic method for extraction of useful information content from time-frequency distributions of nonstationary signals heavily corrupted by additive noise. The proposed method, which does not require prior knowledge of the signal, initially performs a one-dimensional clustering of the time-frequency distribution aimed at segmenting it into a fixed number of classes. This procedure points out basic structural differences of noise and signal components (i.e., useful information) in the time-frequency plane. In fact, noise presents a flat spectrum with large time-frequency supports, while signal components are narrow energy ridges. The time-frequency supports are estimated for each of the obtained classes and used as input for an algorithm aimed at discriminating “noise” affine classes, containing mainly non-zero coefficients located outside the components time-frequency supports, from “useful information” affine classes, containing mainly coefficients inside time-frequency supports of the signal components. Simulations show the superiority of the proposed algorithm, mainly in terms of error rate reduction of classified time-frequency coefficients, compared to hard amplitude thresholding methods, as well as to a recently proposed method for information content extraction based on adaptive thresholding, which is also outperformed in terms of computational complexity and execution time (by up to 43 and 28 times, respectively). [ABSTRACT FROM AUTHOR]

The paper presents an effective solution for improving power system state estimation performance by applying the intersection of confidence intervals (ICI) algorithm, a state-of-the-art adaptive signal processing technique used in signal denoising. Since many power utilities worldwide still run state estimators based on the weighed least squares (WLS) algorithm using supervisory control and data acquisition (SCADA) measurements, the ICI algorithm is added to pre-process SCADA measurements without changing the structure of the WLS algorithm. Due to its adaptive window size and high sensitivity to noise in the input measurement series, the proposed ICI-based solution results in an enhancement of the state estimator output and overall performance when compared to the original algorithm. As test beds, the IEEE systems with 30 and 118 buses were used, while as an example of the real power system, the complete mathematical model of the Croatian transmission power system was simulated. Several case studies indicate that the ICI-based state estimator reduces the input measurements mean squared error by up to 30.8%, the mean absolute error by up to 20.8%, and the maximum estimation error by up to 22.6%. Furthermore, this also led to an enhanced final output of the state estimator for all tested systems in view of the state estimation accuracy and convergence. [ABSTRACT FROM AUTHOR]

Various diseases are diagnosed using medical imaging used for analysing internal anatomical structures. However, medical images are susceptible to noise introduced in both acquisition and transmission processes. We propose an adaptive data-driven image denoising algorithm based on an improvement of the intersection of confidence intervals (ICI), called relative ICI (RICI) algorithm. The 2D mask of the adaptive size and shape is calculated for each image pixel independently, and utilized in the design of the 2D local polynomial approximation (LPA) filters. Denoising performances, in terms of the PSNR, are compared to the original ICI-based method, as well as to the fixed sized filtering. The proposed adaptive RICI-based denoising outperformed the original ICI-based method by up to 1.32 dB, and the fixed size filtering by up to 6.48 dB. Furthermore, since the denoising of each image pixel is done locally and independently, the method is easy to parallelize. [ABSTRACT FROM AUTHOR]

We propose a new algorithm for denoising of additive white Gaussian noise-corrupted signals, based on the intersection of confidence intervals (ICI) algorithm, called the fast intersection of confidence intervals (FICI) algorithm. The proposed approach combines the FICI algorithm, used for the adaptive filter support size selection, with the local polynomial approximation (LPA) method, used as a filter design tool. The LPA-FICI method, when compared to the existing ICI-based denoising method, reduces the computational complexity by up to N times, where N is the number of signal samples, resulting in significantly faster algorithm execution time, while maintaining the estimation accuracy close to the one achieved using the original ICI-based method. Furthermore, the proposed modifications allow the use of the LPA-FICI method in real-time signal processing. In conducted simulations, we have confirmed advantages of the proposed method on two commonly used benchmark signals corrupted with various noise strengths. [ABSTRACT FROM AUTHOR]

In this paper, we study the connections between generalized mean operators and entropies, where the mean value operators are related to the strictly monotone logical operators of fuzzy theory. Here, we propose a new entropy measure based on the family of generalized Dombi operators. Namely, this measure is obtained by using the Dombi operator as a generator function in the general solution of the bisymmetric functional equation. We show how the proposed entropy can be used in a fuzzy system where the performance is consistent in choosing the best alternative in the Multiple Attribute Decision-Making Problem. This newly defined entropy was also applied to the problem of extracting useful information from time-frequency representations of noisy, nonstationary, and multicomponent signals. The denoising results were compared to Shannon and Rényi entropies. The proposed entropy measure is shown to significantly outperform the competing ones in terms of denoising classification accuracy and the F1-score due to its sensitivity to small changes in the probability distribution. [ABSTRACT FROM AUTHOR]

In this paper we have proposed a novel method for image denoising using local polynomial approximation (LPA) combined with the relative intersection of confidence intervals (RICI) rule. The algorithm performs separable column-wise and row-wise image denoising (i.e., independently by rows and by columns), combining the obtained results into the final image estimate. The newly developed method performs competitively among recently published state-of-the-art denoising methods in terms of the peak signal-to-noise ratio (PSNR), even outperforming them for small to medium noise variances for images that are piecewise constant along their rows and columns. [ABSTRACT FROM AUTHOR]

The paper presents an analysis and overview of vertex–frequency analysis, an emerging area in graph signal processing. A strong formal link of this area to classical time–frequency analysis is provided. Vertex–frequency localization-based approaches to analyzing signals on the graph emerged as a response to challenges of analysis of big data on irregular domains. Graph signals are either localized in the vertex domain before the spectral analysis is performed or are localized in the spectral domain prior to the inverse graph Fourier transform is applied. The latter approach is the spectral form of the vertex–frequency analysis, and it will be considered in this paper since the spectral domain for signal localization is well ordered and thus simpler for application to the graph signals. The localized graph Fourier transform is defined based on its counterpart, the short-time Fourier transform, in classical signal analysis. We consider various spectral window forms based on which these transforms can tackle the localized signal behavior. Conditions for the signal reconstruction, known as the overlap-and-add (OLA) and weighted overlap-and-add (WOLA) methods, are also considered. Since the graphs can be very large, the realizations of vertex–frequency representations using polynomial form localization have a particular significance. These forms use only very localized vertex domains, and do not require either the graph Fourier transform or the inverse graph Fourier transform, are computationally efficient. These kinds of implementations are then applied to classical time–frequency analysis since their simplicity can be very attractive for the implementation in the case of large time-domain signals. Spectral varying forms of the localization functions are presented as well. These spectral varying forms are related to the wavelet transform. For completeness, the inversion and signal reconstruction are discussed as well. The presented theory is illustrated and demonstrated on numerical examples. [ABSTRACT FROM AUTHOR]

Electroencephalogram (EEG) signals are known to contain signatures of stimuli that induce brain activities. However, detecting these signatures to classify captured EEG waveforms is one of the most challenging tasks of EEG analysis. This paper proposes a novel time–frequency-based method for EEG analysis and characterization implemented in a computer-aided decision-support system that can be used to assist medical experts in interpreting EEG patterns. The computerized method utilizes EEG spectral non-stationarity, which is clearly revealed in the time–frequency distributions (TFDs) of multicomponent signals. The proposed algorithm, which is based on the modification of the Rényi entropy, called local or short-term Rényi entropy (STRE), was upgraded with a blind component separation procedure and instantaneous frequency (IF) estimation. The method was applied to EEGs of both forward and backward movements of the left and right hands, as well as to EEGs of imagined hand movements, which were captured by a 19-channel EEG recording system. The obtained results show that in a given virtual instrument, the proposed methods efficiently distinguish between real and imagined limb movements by considering their signatures in terms of the dominant EEG component's IFs at the specified subset of EEG channels (namely, F3, F4, F7, F8, T3, and T4). Furthermore, computing the number of EEG signal components, their extraction, and IF estimation provide important information that shows potential to enhance existing clinical diagnostic techniques for detecting the intensity, location, and type of brain function abnormalities in patients with neurological motor control disorders. [ABSTRACT FROM AUTHOR]

Gravitational-wave data (discovered first in 2015 by the Advanced LIGO interferometers and awarded by the Nobel Prize in 2017) are characterized by non-Gaussian and non-stationary noise. The ever-increasing amount of acquired data requires the development of efficient denoising algorithms that will enable the detection of gravitational-wave events embedded in low signal-to-noise-ratio (SNR) environments. In this paper, an algorithm based on the local polynomial approximation (LPA) combined with the relative intersection of confidence intervals (RICI) rule for the filter support selection is proposed to denoise the gravitational-wave burst signals from core collapse supernovae. The LPA-RICI denoising method's performance is tested on three different burst signals, numerically generated and injected into the real-life noise data collected by the Advanced LIGO detector. The analysis of the experimental results obtained by several case studies (conducted at different signal source distances corresponding to the different SNR values) indicates that the LPA-RICI method efficiently removes the noise and simultaneously preserves the morphology of the gravitational-wave burst signals. The technique offers reliable denoising performance even at the very low SNR values. Moreover, the analysis shows that the LPA-RICI method outperforms the approach combining LPA and the original intersection of confidence intervals (ICI) rule, total-variation (TV) based method, the method based on the neighboring thresholding in the short-time Fourier transform (STFT) domain, and three wavelet-based denoising techniques by increasing the improvement in the SNR by up to 118.94% and the peak SNR by up to 138.52%, as well as by reducing the root mean squared error by up to 64.59%, the mean absolute error by up to 55.60%, and the maximum absolute error by up to 84.79%. [ABSTRACT FROM AUTHOR]

The rapid growth in the amount of data in the digital world leads to the need for data compression, and so forth, reducing the number of bits needed to represent a text file, an image, audio, or video content. Compressing data saves storage capacity and speeds up data transmission. In this paper, we focus on the text compression and provide a comparison of algorithms (in particular, entropy-based arithmetic and dictionary-based Lempel–Ziv–Welch (LZW) methods) for text compression in different languages (Croatian, Finnish, Hungarian, Czech, Italian, French, German, and English). The main goal is to answer a question: "How does the language of a text affect the compression ratio?" The results indicated that the compression ratio is affected by the size of the language alphabet, and size or type of the text. For example, The European Green Deal was compressed by 75.79%, 76.17%, 77.33%, 76.84%, 73.25%, 74.63%, 75.14%, and 74.51% using the LZW algorithm, and by 72.54%, 71.47%, 72.87%, 73.43%, 69.62%, 69.94%, 72.42% and 72% using the arithmetic algorithm for the English, German, French, Italian, Czech, Hungarian, Finnish, and Croatian versions, respectively. [ABSTRACT FROM AUTHOR]

Soccer player performance is influenced by multiple unpredictable factors. During a game, score changes and pre-game expectations affect the effort exerted by players. This study used GPS wearable sensors to track players' energy expenditure in 5-min intervals, alongside recording the goal timings and the win and lose probabilities from betting sites. A mathematical model was developed that considers pre-game expectations (e.g., favorite, non-favorite), endurance, and goal difference (GD) dynamics on player effort. Particle Swarm and Nelder–Mead optimization methods were used to construct these models, both consistently converging to similar cost function values. The model outperformed baselines relying solely on mean and median power per GD. This improvement is underscored by the mean absolute error (MAE) of 396.87 ± 61.42 and root mean squared error (RMSE) of 520.69 ± 88.66 achieved by our model, as opposed to the B 1 MAE of 429.04 ± 84.87 and RMSE of 581.34 ± 185.84 , and B 2 MAE of 421.57 ± 95.96 and RMSE of 613.47 ± 300.11 observed across all players in the dataset. This research offers an enhancement to the current approaches for assessing players' responses to contextual factors, particularly GD. By utilizing wearable data and contextual factors, the proposed methods have the potential to improve decision-making and deepen the understanding of individual player characteristics. [ABSTRACT FROM AUTHOR]

• Prospects of sopele woodwind instrument AMT are inspected on a newly acquired dataset. • Unwanted pitch variation is mitigated using DFT and supervised machine learning. • DFT-coupled RF and CNN models achieve F 1 = 0.92 in the polyphonic setup. • A full-stack system for effortless music preservation of sopele pieces is presented. • The system performs reasonably well for transcribing sopele traditional music pieces. Sopela is a traditional hand-made woodwind instrument, commonly played in pair, characteristic to the Istrian peninsula in western Croatia. Its piercing sound, accompanied by two-part singing in the hexatonic Istrian scale, is registered in the UNESCO Representative List of the Intangible Cultural Heritage of Humanity. This paper presents an insight study of automatic music transcription (AMT) for sopele tunes. The process of converting audio inputs into human-readable musical scores involves multi-pitch detection and note tracking. The proposed solution supports this process by utilising frequency-feature extraction, supervised machine learning (ML) algorithms, and postprocessing heuristics. We determined the most favourable tone-predicting model by applying grid search for two state-of-the-art ML techniques, optionally coupled with frequency-feature extraction. The model achieved promising transcription accuracy for both monophonic and polyphonic music sources encompassed in the originally developed dataset. In addition, we developed a proof-of-concept AMT system, comprised of a client mobile application and a server-side API. While the mobile application records, tags and uploads audio sources, the back-end server applies the presented procedure for converting recorded music into a common notation to be delivered as a transcription result. We thus demonstrate how collecting and preserving traditional sopele music, performed in real-life occasions, can be effortlessly accomplished on-the-go. [ABSTRACT FROM AUTHOR]

This paper explores three groups of time–frequency distributions: the Cohen's, affine, and reassigned classes of time–frequency representations (TFRs). This study provides detailed insight into the theory behind the selected TFRs belonging to these classes. Extensive numerical simulations were performed with examples that illustrate the behavior of the analyzed TFR classes in the joint time–frequency domain. The methods were applied both on synthetic and real-life non-stationary signals. The obtained results were assessed with respect to time–frequency concentration (measured by the Rényi entropy), instantaneous frequency (IF) estimation accuracy, cross-term presence in the TFRs, and the computational cost of the TFRs. This study gives valuable insight into the advantages and limitations of the analyzed TFRs and assists in selecting the proper distribution when analyzing given non-stationary signals in the time–frequency domain. [ABSTRACT FROM AUTHOR]

Non-stationary signals are often analyzed using raw waveform data or spectrograms of those data; however, the possibility of alternative time–frequency representations being more informative than the original data or spectrograms is yet to be investigated. This paper tested whether alternative time–frequency representations could be more informative for machine learning classification of seismological data. The mentioned hypothesis was evaluated by training three well-established convolutional neural networks using nine time–frequency representations. The results were compared to the base model, which was trained on the raw waveform data. The signals that were used in the experiment are three-component seismogram instances from the Local Earthquakes and Noise DataBase (LEN-DB). The results demonstrate that Pseudo Wigner–Ville and Wigner–Ville time–frequency representations yield significantly better results than the base model, while spectrogram and Margenau–Hill perform significantly worse (p < 0.01). Interestingly, the spectrogram, which is often used in signal analysis, had inferior performance when compared to the base model. The findings presented in this research could have notable impacts in the fields of geophysics and seismology as the phenomena that were previously hidden in the seismic noise are now more easily identified. Furthermore, the results indicate that applying Pseudo Wigner–Ville or Wigner–Ville time–frequency representations could result in a large increase in earthquakes in the catalogs and lessen the need to add new stations with an overall reduction in the costs. Finally, the proposed approach of extracting valuable information through time–frequency representations could be applied in other domains as well, such as electroencephalogram and electrocardiogram signal analysis, speech recognition, gravitational waves investigation, and so on. [ABSTRACT FROM AUTHOR]

We present a signal decomposition procedure, which separates modes into individual components while preserving their integrity, in effort to tackle the challenges related to the characterization of modes in an acoustic dispersive environment. With this approach, each mode can be analyzed and processed individually, which carries opportunities for new insights into their characterization possibilities. The proposed methodology is based on the eigenanalysis of the autocorrelation matrix of the analyzed signal. When eigenvectors of this matrix are properly linearly combined, each signal component can be separately reconstructed. A proper linear combination is determined based on the minimization of concentration measures calculated exploiting time-frequency representations. In this paper, we engage a steepest-descent-like algorithm for the minimization process. Numerical results support the theory and indicate the applicability of the proposed methodology in the decomposition of acoustic signals in dispersive channels. [ABSTRACT FROM AUTHOR]

In this paper, we present an approach to the reconstruction of signals exhibiting sparsity in a transformation domain, having some heavily disturbed samples. This sparsity-driven signal recovery exploits a carefully suited random sampling consensus (RANSAC) methodology for the selection of a subset of inlier samples. To this aim, two fundamental properties are used: A signal sample represents a linear combination of the sparse coefficients, whereas the disturbance degrades the original signal sparsity. The properly selected samples are further used as measurements in the sparse signal reconstruction, performed using algorithms from the compressive sensing framework. Besides the fact that the disturbance degrades signal sparsity in the transformation domain, no other disturbance-related assumptions are made—there are no special requirements regarding its statistical behavior or the range of its values. As a case study, the discrete Fourier transform is considered as a domain of signal sparsity, owing to its significance in signal processing theory and applications. Numerical results strongly support the presented theory. In addition, the exact relation for the signal-to-noise ratio of the reconstructed signal is also presented. This simple result, which conveniently characterizes the RANSAC-based reconstruction performance, is numerically confirmed by a set of statistical examples. [ABSTRACT FROM AUTHOR]

On-site estimation of sea state parameters is crucial for ship navigation. Extensive research has been conducted on model-based estimation utilizing ship motion responses. Model-free approaches based on machine learning (ML) have recently gained popularity, and estimation from time-series of ship motion responses using deep learning (DL) methods has given promising results. In this study, we apply the novel, attention-based neural network (AT-NN) for estimating wave height, zero-crossing period, and relative wave direction from raw time-series data of ship pitch, heave, and roll. Despite reduced input data, it has been demonstrated that the proposed approaches by modified state-of-the-art techniques (based on convolutional neural networks (CNN) for regression, multivariate long short-term memory CNN, and sliding puzzle neural network) improved estimation MSE, MAE, and NSE by up to 86%, 66%, and 56%, respectively, compared to the best performing original methods for all sea state parameters. Furthermore, the proposed technique based on AT-NN outperformed all tested methods (original and enhanced), improving estimation MSE by 94%, MAE by 74%, and NSE by 80% when considering all sea state parameters. Finally, we proposed a novel approach for interpreting the uncertainty estimation of neural network outputs based on the Monte-Carlo dropout method to enhance the model's trustworthiness. • Enhanced 3 state-of-the-art models with improved estimation of sea state parameters. • Novel attention-based neural network enhancing 3 original and 3 improved models. • Sea state parameters' estimation improved by up to MSE 94%, MSE 74%, and NSE 80%. • Better results achieved utilizing only three ship motion time-series. • Novel way of interpreting the uncertainty estimation of neural network outputs. [ABSTRACT FROM AUTHOR]

The power system state estimator based on the support vector machine (SVM) and the weighted least squares (WLS) method is presented in the paper. The WLS provides state estimations necessary for creating SVM model which is then used for state estimation. The developed algorithm was tested on the IEEE systems, and the performance indicators were calculated in order to compare the accuracy of estimation and the measurement error filtering. The results indicate that the proposed hybrid model outperforms the classical WLS-based state estimation in terms of accuracy and improves measurement error filtering in comparison to the classical estimator. [ABSTRACT FROM AUTHOR]

In clinical practice, fracture age estimation is commonly required, particularly in children with suspected non-accidental injuries. It is usually done by radiologically examining the injured body part and analyzing several indicators of fracture healing such as osteopenia, periosteal reaction, and fracture gap width. However, age-related changes in healing timeframes, inter-individual variabilities in bone density, and significant intra- and inter-operator subjectivity all limit the validity of these radiological clues. To address these issues, for the first time, we suggest an automated neural network-based system for determining the age of a pediatric wrist fracture. In this study, we propose and evaluate a deep learning approach for automatically estimating fracture age. Our dataset included 3570 medical cases with a skewed distribution toward initial consultations. Each medical case includes a lateral and anteroposterior projection of a wrist fracture, as well as patients' age, and gender. We propose a neural network-based system with Monte-Carlo dropout-based uncertainty estimation to address dataset skewness. Furthermore, this research examines how each component of the system contributes to the final forecast and provides an interpretation of different scenarios in system predictions in terms of their uncertainty. The examination of the proposed systems' components showed that the feature-fusion of all available data is necessary to obtain good results. Also, proposing uncertainty estimation in the system increased accuracy and F1-score to a final 0.906 ± 0.011 on a given task. [ABSTRACT FROM AUTHOR]

This paper proposes a new method, ConvGraph, to detect communities in highly cohesive and isolated weighted graphs, where the sum of the weights is significantly higher inside than outside the communities. The method starts by transforming the original graph into a line graph to apply a convolution, a common technique in the computer vision field. Although this technique was originally conceived to detect the optimum edge in images, it is used here to detect the optimum edges in communities identified by their weights rather than by their topology. The method includes a final refinement step applied to communities with a high vertex density that could not be detected in the first phase. The proposed algorithm was tested on a series of highly cohesive and isolated synthetic graphs and on a real-world export graph, performing well in both cases. [ABSTRACT FROM AUTHOR]

Scene classification relying on images is essential in many systems and applications related to remote sensing. The scientific interest in scene classification from remotely collected images is increasing, and many datasets and algorithms are being developed. The introduction of convolutional neural networks (CNN) and other deep learning techniques contributed to vast improvements in the accuracy of image scene classification in such systems. To classify the scene from areal images, we used a two-stream deep architecture. We performed the first part of the classification, the feature extraction, using pre-trained CNN that extracts deep features of aerial images from different network layers: the average pooling layer or some of the previous convolutional layers. Next, we applied feature concatenation on extracted features from various neural networks, after dimensionality reduction was performed on enormous feature vectors. We experimented extensively with different CNN architectures, to get optimal results. Finally, we used the Support Vector Machine (SVM) for the classification of the concatenated features. The competitiveness of the examined technique was evaluated on two real-world datasets: UC Merced and WHU-RS. The obtained classification accuracies demonstrate that the considered method has competitive results compared to other cutting-edge techniques. [ABSTRACT FROM AUTHOR]

The paper proposes a segmentation and classification technique for fracture detection in X-ray images. This novel rotation-invariant method introduces the concept of local entropy for de-noising and removing tissue from the analysed X-ray images, followed by an improved procedure for image segmentation and the detection of regions of interest. The proposed local Shannon entropy was calculated for each image pixel using a sliding 2D window. An initial image segmentation was performed on the entropy representation of the original image. Next, a graph theory-based technique was implemented for the purpose of removing false bone contours and improving the edge detection of long bones. Finally, the paper introduces a classification and localisation procedure for fracture detection by tracking the difference between the extracted contour and the estimation of an ideal healthy one. The proposed hybrid method excels at detecting small fractures (which are hard to detect visually by a radiologist) in the ulna and radius bones—common injuries in children. Therefore, it is imperative that a radiologist inspecting the X-ray image receives a warning from the computerised X-ray analysis system, in order to prevent false-negative diagnoses. The proposed method was applied to a data-set containing 860 X-ray images of child radius and ulna bones (642 fracture-free images and 218 images containing fractures). The obtained results showed the efficiency and robustness of the proposed approach, in terms of segmentation quality and classification accuracy and precision (up to 91.16 % and 86.22 % , respectively). [ABSTRACT FROM AUTHOR]