Search

Combining multiple neural networks has been used to improve the decision accuracy in many application fields including pattern recognition and classification. In this paper, we investigate the potential of this approach for land cover change detection. In a first step, we perform many experiments in order to find the optimal individual networks in terms of architecture and training rule. In the second step, different neural network change detectors are combined using a method based on the notion of fuzzy integral. This method combines objective evidences in the form of network outputs, with subjective measures of their performances. Various forms of the fuzzy integral, which are, namely, Choquet integral, Sugeno integral, and two extensions of Sugeno integral with ordered weighted averaging operators, are implemented. Experimental analysis using error matrices and Kappa analysis showed that the fuzzy integral outperforms individual networks and constitutes an appropriate strategy to increase the accuracy of change detection. [ABSTRACT FROM AUTHOR]

Capsule Neural Network (CapsNet) models are regarded as efficient substitutes for convolutional neural networks (CNN) due to their powerful hierarchical representation capability. Nevertheless, CNN endure their inability of recording spatial information in spectrograms. The main constraint of CapsNet is related to the compression method which can be implemented in CNN models but cannot be directly employed in CapsNet. As a result, we propose a novel architecture based on dual-channel long short-term memory compressed CapsNet (DC-LSTM–COMP CapsNet) for speaker verification in emotional as well as stressful talking environments. The proposed approach is perceived as a modified Capsule network that attempts to overcome the limitations that exist within the original CapsNet, as well as in CNN while enhancing the verification performance. The proposed architecture is assessed on four distinct databases. The experimental analysis reveals that the average speaker verification performance is improved in comparison with CNN, the original CapsNet, as well as the conventional classifiers. The proposed algorithm notably achieves the best verification accuracy across the four speech databases. For example, using the Emirati dataset, the average percentage equal error rates (EERs) obtained is 10.50%, based on the proposed architecture which outperforms other deep and classical models. [ABSTRACT FROM AUTHOR]

This work aims at intensifying text-independent speaker identification performance in real application situations such as noisy and emotional talking conditions. This is achieved by incorporating two different modules: a Computational Auditory Scene Analysis (CASA) based pre-processing module for noise reduction and "cascaded Gaussian Mixture Model – Convolutional Neural Network (GMM-CNN) classifier for speaker identification" followed by emotion recognition. This research proposes and evaluates a novel algorithm to improve the accuracy of speaker identification in emotional and highly-noise susceptible conditions. Experiments demonstrate that the proposed model yields promising results in comparison with other classifiers when "Speech Under Simulated and Actual Stress (SUSAS) database, Emirati Speech Database (ESD), the Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS)" database and the "Fluent Speech Commands" database are used in a noisy environment. • CASA GMM-CNN novel speaker identification is proposed. • The proposed model outperforms GMM-CNN, SVM and MLP. • The proposed algorithm can identify the speaker in noisy conditions. • The proposed model was evaluated on three different datasets. [ABSTRACT FROM AUTHOR]

Application of remotely sensed data in geological mapping has been the subject matter of many studies in the last decades. The optimal exploitation of high spatial resolution satellite imagery can contribute to the improvement of the geological map, particularly when it comes to mapping in arid and desert areas where outcrops are often inaccessible. Djanet area lies in Eastern Hoggar Shield, Algeria. It is characterized by rather high corrugated reliefs and this makes it difficult to collect samples or even get field information from the entire region. The study covers the terrane of Djanet and part of the terrane of Edembo. In addition, this region is probably one of the least explored areas of Hoggar. This is why different remote sensing techniques have been used to make geologic mapping faster and more efficient. Processing of Landsat-7 ETM+ images covering the Djanet region includes various treatments ranging from contrast enhancements to spectral enhancements, and using images in 742 RGB color compositions, calculation processing of band ratios (3/1, 5/4, 7/5) and the (5/7, 2/1, 4/2) and the application of the directional filters, according to the orientations ranging from N000 up to N180E, with a semi-automatic approach for the extraction of lineaments to establish two maps, a lithological map and another linear map. The correlation of these maps with the field data allowed us to check the validity and the correspondence of the different facies and especially to clarify the lithological contours and the identification of the majority of the tectonic accidents derived from the analysis of Landsat satellite data. The geological map obtained gives another image of the region studied and provides new information on the contacts of the lithological units and the structural diagram. [ABSTRACT FROM AUTHOR]

• An approved method based on the RLS algorithm is presented. • Comparison between the standard RLS algorithm and the proposed method. • The proposed method performances are confirmed by experiment tests. This paper describes another way to perform the on-line induction machine's parameter identification using the standard linear form of the recursive least-squares algorithm (RLS), which is commonly applied to estimate the machine parameters in motoring mode using an appropriate linear form machine model. The proposed approach shows that with this same methodology, better results could be obtained when the considered machine is running in self-excited generating mode. The on-line parameter estimation task is performed just during the beginning of the voltage build-up process where the magnetizing machine curve is linear and both saturated and non-saturated machine models present the same dynamic behavior. Since the rotor is driven at a constant speed and the required reactive power is provided by a capacitors bank connected to the stator windings terminals, this fact allows a direct application of the standard RLS estimation algorithm to the frequently linear machine model without any voltage measurements and without any additional restrictions generally imposed by the motoring mode. These potential advantages will contribute to significantly reduce the estimation algorithm complexity. Simulation results validated by experimental tests confirm the effectiveness of the proposed approach with improved accuracy. [ABSTRACT FROM AUTHOR]

• The paper focuses on writer retrieval in handwritten document datasets. • New combination scheme based on dissimilarity learning is proposed. • New features based on multi-scale Histogram of Templates are proposed. • Introduce Multi-gradient Elongated Quinary Patterns for handwriting characterization. • The combination results outperform the state of the art. Retrieval based on query images supports interesting applications in handwritten document analysis, such as checking manuscripts originality, and authorship. In this respect, writer retrieval systems aim to automatically find all manuscripts belonging to the same author. Presently, we propose a new combination scheme for multiple writer retrieval systems that employ different features and dissimilarities. The proposed combination is founded on writer-independent, SVM dissimilarity learning. For experimental evaluation, three individual systems are proposed each of which, has its specific features. To develop the first system, we propose the Multiscale Histogram Of Templates (M-HOT). For the second system, we introduce the so-called Multi-Gradient Elongated Quinary Pattern (MG-EQP) as new descriptor for handwriting characterization. The third system uses the well-known Run Length Features. Retrieval tests are performed on CVL, ICDAR-2011, ICDAR-2013 and ICDAR-2017 datasets. Furthermore, to highlight the language-independence aspect, experiments are performed on KHATT dataset that contains Arabic handwritten documents. Results obtained evince the effectiveness of the proposed features as well as the combination scheme, which outperforms both individual systems and the state of the art. [ABSTRACT FROM AUTHOR]

In this work, we conducted an empirical comparative study of the performance of text-independent speaker verification in emotional and stressful environments. This work combined deep models with shallow architecture, which resulted in novel hybrid classifiers. Four distinct hybrid models were utilized: deep neural network-hidden Markov model (DNN-HMM), deep neural network-Gaussian mixture model (DNN-GMM), Gaussian mixture model-deep neural network (GMM-DNN), and hidden Markov model-deep neural network (HMM-DNN). All models were based on novel implemented architecture. The comparative study used three distinct speech datasets: a private Arabic dataset and two public English databases, namely Speech Under Simulated and Actual Stress (SUSAS) and Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS). The test results of the aforementioned hybrid models demonstrated that the proposed HMM-DNN leveraged the verification performance in emotional and stressful environments. Results also showed that HMM-DNN outperformed all other hybrid models in terms of equal error rate (EER) and area under the curve (AUC) evaluation metrics. The average resulting verification system based on the three datasets yielded EERs of 7.19, 16.85, 11.51, and 11.90% based on HMM-DNN, DNN-HMM, DNN-GMM, and GMM-DNN, respectively. Furthermore, we found that the DNN-GMM model demonstrated the least computational complexity compared to all other hybrid models in both talking environments. Conversely, the HMM-DNN model required the greatest amount of training time. Findings also demonstrated that EER and AUC values depended on the database when comparing average emotional and stressful performances. [ABSTRACT FROM AUTHOR]

One-class classifier (OCC) is involved for solving different kinds of problems due to its ability to represent a class distribution regardless the remaining classes. Its main advantage for multi-class classification is offering an open system and therefore allows easily extending new classes without retraining OCCs. So far, hidden Markov models, support vector machines and neural networks are the most used classifiers for Arabic word recognition, which provides a system with closed lexicon. In this paper, the OCCs are explored in order to perform an Arabic word recognition system with an open lexicon. Generally, pattern recognition systems designed by a single system suffer from limitations such as the lack of uniqueness and non-universality. Thus, combining multiple systems becomes an attractive research topic for performance and robustness enhancement. Fixed rules are commonly used us combiners for the hybrid OCC ensembles. The present paper aims to propose a combination scheme of OCCs based on the use of fuzzy integral (FI) operators. Furthermore, an alternative framework is proposed to design a parameter-independent and open-lexicon handwritten Arabic word recognition system as well as a new density measure function. Experimental results conducted on Arabic handwritten dataset using different types of OCCs with large number of classes highlight the superiority of FI for hybrid OCC ensembles. [ABSTRACT FROM AUTHOR]

Speech signals carry various bits of information relevant to the speaker such as age, gender, accent, language, health, and emotions. Emotions are conveyed through modulations of facial and vocal expressions. This paper conducts an empirical comparison of performances between the classical classifiers: Gaussian Mixture Model (GMM), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Artificial neural networks (ANN); and the deep learning classifiers, i.e., Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and Gated Recurrent Unit (GRU) in addition to the ivector approach for a text-independent speaker verification task in neutral and emotional talking environments. The deep models undergo hyperparameter tuning using the Grid Search optimization algorithm. The models are trained and tested using a private Arabic Emirati Speech Database, Ryerson Audio–Visual Database of Emotional Speech and Song dataset (RAVDESS) database, and a public Crowd-Sourced Emotional Multimodal Actors (CREMA) database. Experimental results illustrate that deep architectures do not necessarily outperform classical classifiers. In fact, evaluation was carried out through Equal Error Rate (EER) along with Area Under the Curve (AUC) scores. The findings reveal that the GMM model yields the lowest EER values and the best AUC scores across all datasets, amongst classical classifiers. In addition, the ivector model surpasses all the fine-tuned deep models (CNN, LSTM, and GRU) based on both evaluation metrics in the neutral, as well as the emotional speech. In addition, the GMM outperforms the ivector using the Emirati and RAVDESS databases. [ABSTRACT FROM AUTHOR]

Objectives: The aim of this study was to investigate the value of metabolic tumor imaging using hybrid PET for the preoperative detection of extranodal extension (ENE) in lymph node metastases of oropharyngeal squamous cell carcinoma (OPSCC)., Methods: We performed a retrospective analysis of a consecutive cohort of patients with OPSCC treated with primary surgery with or without adjuvant (chemo-) radiotherapy at the Kantonsspital Sankt-Gallen and the University Hospital Zurich, Switzerland, from 2010 until 2019. Hybrid PET was compared to conventional cross-sectional imaging with MRI and CT. Histopathological presence of ENE of neck dissection specimen served as gold standard., Results: A total number of 234 patients were included in the study, 95 (40.6%) of which had pathological ENE (pENE). CT has a good specificity with 93.7%; meanwhile, MRI was the most sensitive diagnostic method (72.0%). The nodal metabolic tumor parameters (SUV max , TLG, MTV) were significantly higher in patients with positive ENE (p < 0.001 for all three parameters) than in patients with negative ENE (p < 0.001, for all three parameters)., Conclusions: CT achieved the best specificity, while MRI had the best sensitivity to detect ENE. Nodal metabolic tumor parameters differed significantly between ENE-positive/negative and p16-positive/negative patients. Hence, quantitative data obtained by metabolic imaging might predict presence of ENE and, therefore, could be helpful in customizing therapy management., (© 2022. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)

This paper presents new results on the gasification of spent pot lining (SPL) waste material generated in the primary aluminium smelting industry. The main objective is to test the performance of the gasification process of treated SPL materials and to develop an optimization method to maximize the quality of syngas fuel. The novelty of this study is the development of statistical models to predict the syngas composition and the gasification performance indicators during the SPL waste materials thermal conversion process. Modelling and simulation analysis are performed to convert the SPL solid materials to syngas fuel. The percentage of hydrogen (H 2 ) and carbon monoxide (CO) in the syngas fuel, the cold gasification efficiency (CGE) and the carbon conversion (CC) are determined. The response surface methodology (RSM) is used for the optimization of the performance of the gasification process. The effects of the input factors such as the temperature, the equivalence ratio and the steam to fuel ratio on the output variables (H 2 and CO in the syngas, the CGE and the CC) are determined. The optimization results show that the optimized operating parameters to maximize the H 2 , CO, CGE and CC were T = 1200 °C, ER = 0.1 and SFR = 1.29, respectively. The optimum values for the H 2 , CO, CGE and CC were 37.2%, 22.2%, 79.75% and 97.7%, respectively. New correlations for the variation of the output variables versus the input factors are also presented., (© 2022. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)