Decision Fusion

The world is facing increasing land scarcity due to growing demand for agricultural products and urban expansion. At the same time, farmland abandonment is emerging as a widespread global land-use change phenomenon. Remote sensing plays a critical role in identifying abandonment across diverse farming systems. Here, we synthesized current knowledge through a systematic literature review of 131 publications to assess progress in remote-sensing-based monitoring of farmland abandonment. Our review highlights the growing use of remote sensing techniques and the increasing utility of multisource satellite data. However, research remains primarily skewed toward publicly available optical Landsat and Sentinel-2 data, with limited integration of other sources and a lack of global-scale assessments. We propose research directions to guide future studies, focusing on underrepresented land-use types, such as grasslands, terraces, and plantations, as well as regions like Africa, Central and Southeast Asia, and South America. We emphasize the importance of diversifying data sources and integrating ancillary information, including cadastral data and historical land-use records, to better understand abandonment processes and associated vegetation changes. We advocate for multi-scale, temporally explicit analyses to improve scalability and enable the development of regional and continental products. As most studies focus on biophysical changes, future work should also consider socio-economic contexts and integrate remotesensing-based proxies of land-use change. Finally, we recommend improving communication by clearly defining abandonment, providing visual examples, validating with diverse reference data, documenting uncertainty, and sharing data, code, and outputs.

The 2013 Data Fusion Contest organized by the Data Fusion Technical Committee (DFTC) of the IEEE Geoscience and Remote Sensing Society aimed at investigating the synergistic use of hyperspectral and Light Detection And Ranging (LiDAR) data. The data sets distributed to the participants during the Contest, a hyperspectral imagery and the corresponding LiDAR-derived digital surface model (DSM), were acquired by the NSF-funded Center for Airborne Laser Mapping over the University of Houston campus and its neighboring area in the summer of 2012. This paper highlights the two awarded research contributions, which investigated different approaches for the fusion of hyperspectral and LiDAR data, including a combined unsupervised and supervised classification scheme, and a graph-based method for the fusion of spectral, spatial, and elevation information.

In this paper, we address the multimodal biometric decision fusion problem. By exploring into the user-specific approach for learning and threshold setting, four possible paradigms for learning and decision making are investigated. Since each user requires a decision hyperplane specific to him in order to achieve good verification accuracy, those tedious iterative training methods like the neural network approach would not be suitable. We propose to use a model that requires only a single training step for this application. The four global and local learning and decision paradigms are then explored to observe their decision capability. Besides the proposal of a relevant receiver operating characteristic performance for the local decision, extensive experiments were conducted to observe the verification performance for fusion of two and three biometrics.

This paper describes a process for aggregating different information sources to estimate remaining equipment life. Specifically, the approach presents a rigorous chain of preprocessing, modeling and postprocessing steps that arrive at the desired prognostic result. The preprocessing steps deal with data reduction, filtering, and signature amplification. The prediction model applies Adaptive Neuro-Fuzzy Inference System (ANFIS) to the data. The post-processing steps include recursive trending which implicitly forces the prognostic trend to be confirmed before updated estimates are reported. Prognostic false positives and false negatives are introduced as innovative measures that help in assessing the performance of the approach. The method is illustrated using real-life data from industrial web paper breakage prediction.

Technologies (ICT) has increased privacy concerns among citizens. In this paper we examine how interconnected modules could be developed to mitigate those concerns and foster trust among e-Government stakeholders. Current e-Government measurement models do not adequately address privacy issues. We suggest that privacy enforcements should be added as major criteria of evaluation. Moreover, users should be able to choose the level of privacy and security needed for any system interaction. Thus, a conceptual model that allows for users to easily control the selection of privacy preferences is needed. A Decision Fusion Module takes the user inputs, such as level of privacy, pseudonymity and trust, and processes them to output the best possible selection of security mechanisms corresponding to those user preferences. In fact, e-Government models could benefit from the earned user trust.

Tuberculosis is a major health threat in many regions of the world. When left undiagnosed and thus untreated, mortality rates of patients with tuberculosis are high. Initially, remove the noise from the images. For filtering the images the median filter is used for denoising. In a second step, "Graph cuts" is applied specifically to the filtered models which employ a max flow / min-cut optimization. After lung segmentation seven features are extracted. Local Binary Pattern (LBP) is a simple yet very efficient texture operator which labels the pixels of an image by thresholding the neighborhood of each pixel and considers the result as a binary number. HOG counts occurrences of gradient orientation in localized portions of an image. It also computed on a dense grid of uniformly spaced cells and uses overlapping local contrast normalization for improved accuracy. The Tamura feature is extracted which is motivated by the human visual perception. And the Statistical pattern feature extraction and Gray level co-occurrence matrix are extracted followed by SASI Feature and PCA Feature. These features are given to the RVM classifier. An RVM model is a representation of the examples as points in space, mapped so that the examples of the separate categories are divided by a clear gap that is as wide as possible. The trained classifier will predict about the CXR lung images. And finally the classifier performance is analyzed with the existing system.

Decision fusion is one of hot research topics in classification area, which aims to achieve the best possible performance for the task at hand. In this paper, we investigate the usefulness of this concept to improve change detection accuracy in remote sensing. Thereby, outputs of two fuzzy change detectors based respectively on simultaneous and comparative analysis of multitemporal data are fused by using fuzzy integral operators. This method fuses the objective evidences produced by the change detectors with respect to fuzzy measures that express the difference of performance between them. The proposed fusion framework is evaluated in comparison with some ordinary fuzzy aggregation operators. Experiments carried out on two SPOT images showed that the fuzzy integral was the best performing. It improves the change detection accuracy while attempting to equalize the accuracy rate in both change and no change classes.

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

My deepest gratitude goes to my advisors Lorenzo and Jón. My reaching is also due to their insightful advice, their many last-minute corrections and their flexibility in coping with the necessities due to the growth of my family. Over almost four years, a solid professional relationship, not to mention a friendship has been built up. And this has further enriched the experience. Moving down the rank of the professional hierarchical tree, I would like to thank the RSLab, in particular the lasting members (listed in alphabetical order): Adamo, Begüm, Claudio, Francesca, Luca, Michele, Silvia (a special thank you to her for having shared a desk with me for these three years) and Patra. We have been in each other's company on a three-days-per-week basis. Concerning my stays at 64.139N, 021.953W, Iceland (an awesome country!), I must mention Alberto, Giulia, Hildur, Nicola and Gísli, Róbert from Oxymap. A particular thanks goes to Alberto for having paid me several outstanding rúsínur-included dinners. Thanks also to Hjörtur and the italo-icelandic community of Laugardalslaug (Ester, Leone and Roberto) for their company during my stays. How could I forget Sundhöll, Laugardalslaug and Vesturbaejarlaug (my favorite!) for the many happy hours spent there. I also wish to thank prof. Philippe Salembier and prof. Paolo Gamba (for providing the data sets of Pavia) for being part of the committee. Now that the academic acknowledgments are over, I thank Alberto (l'Albe), for the mid-day routine: pizza and discussions on a lot of low-level technical stuff. Thanks also to Nicola (il Nico), the abstract discussions on the multivariate analysis of the fMRI data have accompanied me on many return trips to Verona. I also take this opportunity to mention my friends: Giana, Davide, Cic, Iulbi, Claudio and the crew of the swimming pool of Fumane, for having shared the few moments of spare time available to me. Thanks go to my mother Carla and my sister Marta (and Alessandro) especially for their considerable help with the kids. The work on innumerable occasions has been helped by this. Last in this list but first for importance, I want to thank my familyteam: Francesca, Alessandro and Massimo. Without whom I would have done much more, but every thing would be much less significant.

Morphological profiles (MPs) are a useful tool for remotely sensed image classification. These profiles are constructed on a base image that can be a single band of a multicomponent remote sensing image. Principal component analysis (PCA) has been used to provide other base images to construct MPs in high-dimensional remote sensing scenes such as hyperspectral images [e.g., by deriving the first principal components (PCs) and building the MPs on the first few components]. In this paper, we discuss several strategies for producing the base images for MPs, and further categorize the considered methods into four classes: 1) linear, 2) nonlinear, 3) manifold learning-based, and 4) multilinear transformation-based. It is found that the multilinear PCA (MPCA) is a powerful approach for base image extraction. That is because it is a tensor-based feature representation approach, which is able to simultaneously exploit the spectralspatial correlation between neighboring pixels. We also show that independent component analysis (ICA) is more effective for constructing base images than PCA. Another important contribution of this paper is a new concept of multiple MPs (MMPs), aimed at synthesizing the spectral-spatial information extracted from the multicomponent base images, and further enhancing the classification accuracy of MPs. Moreover, we propose two different strategies to interpret the newly proposed MMPs by considering their hyperdimensional feature space: 1) decision fusion and 2) sparse classifier based on multinomial logistic regression (MLR). Experiments conducted on three well-known

Extended Attribute Profiles (EAPs), which are obtained by applying morphological attribute filters to an image in a multilevel architecture, can be used for the characterization of the spatial characteristics of objects in a scene. EAPs have proved to be discriminant features when considered for thematic classification in remote sensing applications especially when dealing with very high resolution images. Altimeter data (such as LiDAR) can provide important information, which being complementary to the spectral one can be valuable for a better characterization of the surveyed scene. In this paper, we propose a technique performing a classification of the features extracted with EAPs computed on both optical and LiDAR images, leading to a fusion of the spectral, spatial and elevation data. The experiments were carried out on LiDAR data along either with a hyperspectral and a multispectral image acquired on a rural and urban area of the city of Trento (Italy), respectively. The classification accuracies obtained pointed out the effectiveness of the features extracted by EAPs on both optical and LiDAR data for classification.

My deepest gratitude goes to my advisors Lorenzo and Jón. My reaching is also due to their insightful advice, their many last-minute corrections and their flexibility in coping with the necessities due to the growth of my family. Over almost four years, a solid professional relationship, not to mention a friendship has been built up. And this has further enriched the experience. Moving down the rank of the professional hierarchical tree, I would like to thank the RSLab, in particular the lasting members (listed in alphabetical order): Adamo, Begüm, Claudio, Francesca, Luca, Michele, Silvia (a special thank you to her for having shared a desk with me for these three years) and Patra. We have been in each other's company on a three-days-per-week basis. Concerning my stays at 64.139N, 021.953W, Iceland (an awesome country!), I must mention Alberto, Giulia, Hildur, Nicola and Gísli, Róbert from Oxymap. A particular thanks goes to Alberto for having paid me several outstanding rúsínur-included dinners. Thanks also to Hjörtur and the italo-icelandic community of Laugardalslaug (Ester, Leone and Roberto) for their company during my stays. How could I forget Sundhöll, Laugardalslaug and Vesturbaejarlaug (my favorite!) for the many happy hours spent there. I also wish to thank prof. Philippe Salembier and prof. Paolo Gamba (for providing the data sets of Pavia) for being part of the committee. Now that the academic acknowledgments are over, I thank Alberto (l'Albe), for the mid-day routine: pizza and discussions on a lot of low-level technical stuff. Thanks also to Nicola (il Nico), the abstract discussions on the multivariate analysis of the fMRI data have accompanied me on many return trips to Verona. I also take this opportunity to mention my friends: Giana, Davide, Cic, Iulbi, Claudio and the crew of the swimming pool of Fumane, for having shared the few moments of spare time available to me. Thanks go to my mother Carla and my sister Marta (and Alessandro) especially for their considerable help with the kids. The work on innumerable occasions has been helped by this. Last in this list but first for importance, I want to thank my familyteam: Francesca, Alessandro and Massimo. Without whom I would have done much more, but every thing would be much less significant.

Working mothers are more common in modern society. Child care has consequently turned into a daily struggle for many households. As a result, the majority of parents drop off their kids at a daycare facility or at their grandparents' home. Parents, however, are unable to watch over their kids continuously, whether things are normal or not. A cost-effective IoTbased solution for real-time monitoring is therefore suggested: the Internet of Things-based baby monitoring system (IoT-BBMS). Additionally, we have suggested a new algorithm for our system that is crucial in delivering better child care when parents are gone. In the system automatic music system will be activated as soon as the baby's crying level is detected while the baby is in the swing, which will help stop the baby's crying. The prototype of the proposed system has been developed and tested to prove its effectiveness in terms of cost and simplicity and to ensure secure operation to enable child parenting anywhere and anytime over the network.

In this paper we propose a robust method for face recognition using a descriptor which has been designed to search for features using a grid pattern. This is different from popular key point descriptors which search images only in positions where they expect to extract the most features. An image segmentation technique to modify the representations of images of some standard datasets by enhancing its salient features. The descriptor provides us with a vector of key points of the image. The features obtained from the keypoints that are extracted are aggregated using a feature aggregator. A comparison study on the results obtained from classification of these aggregated features using different non-linear machine learning classifiers is performed.

In this paper we propose a robust method for face recognition using a descriptor which has been designed to search for features using a grid pattern. This is different from popular key point descriptors which search images only in positions where they expect to extract the most features. An image segmentation technique to modify the representations of images of some standard datasets by enhancing its salient features. The descriptor provides us with a vector of key points of the image. The features obtained from the keypoints that are extracted are aggregated using a feature aggregator. A comparison study on the results obtained from classification of these aggregated features using different non-linear machine learning classifiers is performed.

The Cognitive Radio (CR) concept aims to opportunistically access the different bandwidths of wireless spectrum with preventing harmful interference to the licensed users to address the fixed bands access issue. The Spectrum Sensing (SS) stage is crucial in CR system to reliably estimate the presenting of licensed user. This stage can be performed thorough several techniques to reuse the OFDM signal in 4G and 5G systems. The Energy Detection (ED) technique is the simplest and low computational complexity. However, it has some challenges in low Signal-to-Noise Ratio (SNR). In this thesis, we address the challenges in addition to reduce the computational complexity and improve the detection accuracy. Firstly, a new approximated closed-form expression is derived for the non-cooperative SS (NSS) based on ED technique to sense an OFDM signal. With the purpose of low SNR effect reduction, the expression is presented a novel Constant Local False Alarm Rate (CLFAR) with Constant Local Detec...

The goal of feature induction is to automatically create nonlinear combinations of existing features as additional input features to improve classification accuracy. Typically, nonlinear features are introduced into a support vector machine (SVM) through a nonlinear kernel function. One disadvantage of such an approach is that the feature space induced by a kernel function is usually of high dimension and therefore will substantially increase the chance of over-fitting the training data. Another disadvantage is that nonlinear features are induced implicitly and therefore are difficult for people to understand which induced features are critical to the classification performance. In this paper, we propose a boosting-style algorithm that can explicitly induces important nonlinear features for SVMs. We present empirical studies with discussion to show that this approach is effective in improving classification accuracy for SVMs. The comparison with an SVM model using nonlinear kernels also indicates that this approach is effective and robust, particularly when the number of training data is small.

In this paper we propose a novel approach to image compression based on three-dimensional Discrete Cosine Transformation (DCT). The basic idea is to de-correlate similar pixel blocks through three-dimensional DCT transformation. A number of adjacent pixel blocks are grouped together to form a three -dimensional data cube. Each data cube is 3D DCT transformed, quantized, and Huffman encoded. Experimental results demonstrate the effectiveness of the new approach, specifically for medical and space exploration images.

Multiclass image classification is a complex task that has been thoroughly investigated in the past. Decomposition-based strategies are commonly employed to address it. Typically, these methods divide the original problem into smaller, potentially simpler problems, allowing the application of numerous well-established learning algorithms that may not apply directly to the original task. This work focuses on the efficiency of decomposition-based methods and proposes several improvements to the meta-learning level. In this paper, four methods for optimizing the ensemble phase of multiclass classification are introduced. The first demonstrates that employing a mixture of experts scheme can drastically reduce the number of operations in the training phase by eliminating redundant learning processes in decomposition-based techniques for multiclass problems. The second technique for combining learner-based outcomes relies on Bayes’ theorem. Combining the Bayes rule with arbitrary decompos...

This paper propose two decision fusion-based multitemporal classifiers, namely, the jointly likelihood and the weighted majority fusion classifiers, that are derived using two different definitions of the minimum expected cost. Without any overhead incurred by multitemporal processing, a user-selected conventional pixelwise classifier makes local class decisions separately using each temporal data set, and the proposed multitemporal classifiers make the global class decisions by optimally summarizing those local class decisions. The weighted majority decision fusion classifier can handle not only the data set reliabilities but also the classwise reliabilities of each data set. Classification experiment using the jointly likelihood decision fusion with three remotely sensed Thematic Mapper (TM) data sets shows more than 10% overall classification accuracy improvement over the pixelwise maximum likelihood classifier.

Two types of measurements are usually performed from high resolution ECG recordings: (a) static parameters derived from the signal-averaged QRS complex and (b) variant markers derived from beat-to-beat recordings. It is known that an increased QRS micro-variability and ventricular late potentials are associated with an increased risk for malignant arrhythmias. However, the diagnostic power of the singular parameters is limited In this study we investigated the diagnostic ability of a decision fusion of both variant and static high-resolution ECG parameters with radial-basis-function (RBF) networks. Continuous and signal-averaged ECGs were recorded from 51 healthy volunteers without any structural heart disease and no cardiac risk factors and from 44 patients with coronary heart disease and ventricular arrhythmias. Beat-to-beat micro-variability measurement of the QRS complex and the ST-T segment was based on 250 consecutive sinus beats per individual. Signal-averaged ECGs were analy...

This work investigates Distributed Detection (DD) in Wireless Sensor Networks (WSNs) utilizing channel-aware binary-decision fusion over a shared flat-fading channel. A reconfigurable metasurface, positioned in the near-field of a limited number of receive antennas, is integrated to enable a holographic Decision Fusion (DF) system. This approach minimizes the need for multiple RF chains while leveraging the benefits of a large array. The optimal fusion rule for a fixed metasurface configuration is derived, alongside two suboptimal joint fusion rule and metasurface design strategies. These suboptimal approaches strike a balance between reduced complexity and lower system knowledge requirements, making them practical alternatives. The design objective focuses on effectively conveying the information regarding the phenomenon of interest to the FC while promoting energy-efficient data analytics aligned with the Internet of Things (IoT) paradigm. Simulation results underscore the viability of holographic DF, demonstrating its advantages even with suboptimal designs and highlighting the significant energy-efficiency gains achieved by the proposed system.

In this paper, a technique based on Independent Component Analysis (ICA) and morphological attribute filters is presented for the classification of high geometrical resolution hyperspectral images. The ICA is computed instead of the conventional principal component analysis (PCA) in order to better model the information in the hyperspectral image. The spatial characteristics of the objects in the scene are modeled by different multi-level attribute filters. Moreover, a method for increasing the robustness of the analysis based on a decision fusion strategy is proposed. A hyperspectral high resolution image acquired over the city of Pavia (Italy) was considered in the experiments.

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

In speaker verification, a claimant may produce two or more utterances. In our previous study , we proposed to compute the optimal weights for fusing the scores of these utterances based on their score distribution and our prior knowledge about the score statistics estimated from the mean scores of the corresponding client speaker and some pseudo-impostors during enrollment. As the fusion weights depend on the prior scores, in this paper, we propose to adapt the prior scores during verification based on the likelihood of the claimant being an impostor. To this end, a pseudo-imposter GMM score model is created for each speaker. During verification, the claimant's scores are fed to the score model to obtain a likelihood for adapting the prior score. Experimental results based on the GSM-transcoded speech of 150 speakers from the HTIMIT corpus demonstrate that the proposed prior score adaptation approach provides a relative error reduction of 15% when compared with our previous approach where the prior scores are non-adaptive.

In speaker verification, a claimant may produce two or more utterances. Typically, the scores of the speech patterns extracted from these utterances are averaged and the resulting mean score is compared with a decision threshold. Rather than simply computing the mean score, we propose to compute the optimal weights for fusing the scores based on the score distribution of the independent utterances and our prior knowledge about the score statistics. More specifically, we use enrollment data to compute the mean scores of client speakers and impostors and consider them to be the prior scores. During verification, we set the fusion weights for individual speech patterns to be a function of the dispersion between the scores of these speech patterns and the prior scores. Experimental results based on the GSM-transcoded speech of 150 speakers from the HTIMIT corpus demonstrate that the proposed fusion algorithm can increase the dispersion between the mean speaker scores and the mean impostor scores. Compared with a baseline approach where equal weights are assigned to all scores, the proposed approach provides a relative error reduction of 19%.

Fusion techniques have been widely used in multi-modal biometric authentication systems. While these techniques are mainly applied to combine the outputs of modality-dependent classifiers, they can also be applied to fuse the decisions or scores from a single modality. The idea is to consider the multiple samples extracted from a single modality as independent but coming from the same source. In this chapter, we propose a single-source, multi-sample data-dependent fusion algorithm for speaker verification. The algorithm is data-dependent in that the fusion weights are dependent on the verification scores and the prior score statistics of claimed speakers and background speakers. To obtain the best out of the speaker's scores, scores from multiple utterances are sorted before they are probabilistically combined. Evaluations based on 150 speakers from a GSM-transcoded corpus are presented. Results show that data-dependent fusion of speaker's scores is significantly better than the conventional score averaging approach. It was also found that the proposed fusion algorithm can be further enhanced by sorting the score sequences before they are probabilistically combined.

Infant care remains a critical concern for parents and caregivers, requiring constant supervision to ensure the child's safety and well-being. Traditional baby monitoring systems primarily rely on audio and video surveillance, necessitating continuous human attention and lacking proactive response mechanisms. These conventional solutions are often inefficient in handling emergency situations, as they do not incorporate predictive analytics or automated intervention capabilities. This paper presents an AI and IoT-integrated Smart Baby Monitoring System that leverages deep learning for infant emotion and cry detection, IoT-enabled environmental sensing, and automated cradle actuation for enhanced infant care. The proposed system employs an ATmega328 microcontroller for hardware control, an ESP8266 Wi-Fi module for cloud connectivity, and a ThingSpeak-based cloud analytics platform for real-time monitoring and remote accessibility. The AI model is trained to classify infant cries and emotional states, distinguishing between hunger, discomfort, and distress, enabling prompt and context-aware responses. Additionally, environmental sensors continuously track temperature and humidity, ensuring a safe sleeping environment. Automated intervention mechanisms, such as cradle movement and lullaby playback, provide immediate soothing actions to reduce caregiver dependency. Experimental evaluations demonstrate significant improvements in detection accuracy, response efficiency, and energy optimization compared to conventional baby monitoring approaches. The findings highlight the potential of AI-driven predictive analytics in enhancing infant safety, reducing caregiver stress, and improving overall monitoring efficiency in real-world applications.

The problem of decision fusion in distributed sensor systems is considered. Distributed sensors pass their decisions to a fusion center that combines the received decisions from the various sensors into a final global decision. The case where only two sensors are combined using an AND fusion rule was analyzed by Kovattana (1973), and the cases where two and three sensors

Consider a sensor network made of remote nodes connected to a common fusion center. In a recent work Blum and Sadler [1] propose the idea of ordered transmissions -sensors with more informative samples deliver their messages first-and prove that optimal detection performance can be achieved using only a subset of the total messages. Taking to one extreme this approach, we show that just a single delivering allows making the detection errors as small as desired, for a sufficiently large network size: a one-bit detection scheme can be asymptotically consistent. The transmission ordering is based on the modulus of some local statistic (MO system). We derive analytical results proving the asymptotic consistency and, for the particular case that the local statistic is the log-likelihood (ℓ-MO system), we also obtain a bound on the error convergence rate. All the theorems are proved under the general setup of random number of sensors. Poisson-deployed networks, detection by per-sensor censoring, monitoring of energy-constrained phenomenon.

Individual classifiers that are fully trained are unstable especially when the database conditions are changed. Moreover, designing a unique classifier with the suitable parameters to achieve acceptable performance is a non-trivial task. Combined classifiers, which consist of a set of individually trained classifiers, are introduced to avoid the previous problems. There are two key issues in the combination of classifiers. The first issue is how to obtain the set of base classifiers to combine. The second issue is how to fuse the decisions of those classifiers. In this paper, weak Learning Vector Quantization (LVQ) neural networks have been used as base classifiers. Also, a new combination technique which is based on training-weighted voting is introduced. Other factors that greatly affect the performance of a combined classifier are related to the type of the individual classifiers, the training parameters, database size and nature, etc. These factors have been considered in the design of the proposed combined classifier. TWE has been experimentally tested on five standard face databases: Yale, ORL, Grimace, Faces94 and Faces95 and has demonstrated excellent performance. Analysis of the ensemble stability has shown promising results.

In today's competitive landscape, businesses grapple with customer retention. Churn prediction models, although beneficial, often lack accuracy due to the reliance on a single data source. The intricate nature of human behavior and highdimensional customer data further complicate these efforts. To address these concerns, this paper proposes a multimodal fusion learning model for identifying customer churn risk levels in financial service providers. Our multimodal approach integrates customer sentiments, financial literacy (FL) level, and financial behavioral data, enabling more accurate and bias-free churn prediction models. The proposed FL model utilizes a SMOGN-COREG supervised model to gauge customer FL levels from their financial data. The baseline churn model applies an ensemble artificial neural network and oversampling techniques to predict churn propensity in high-dimensional financial data. We also incorporate a speech emotion recognition model employing a pretrained CNN-VGG16 to recognize customer emotions based on pitch, energy, and tone. To integrate these diverse features while retaining unique insights, we introduced late and hybrid fusion techniques that complementary boost coordinated multimodal colearning. Robust metrics were utilized to evaluate the proposed multimodal fusion model and hence the approach's validity, including mean average precision and macro-averaged F1 score. Our novel approach demonstrates a marked improvement in churn prediction, achieving a test accuracy of 91.2%, a Mean Average Precision (MAP) score of 66, and a Macro-Averaged F1 score of 54 through the proposed hybrid fusion learning technique compared with late fusion and baseline models. Furthermore, the analysis demonstrates a positive correlation between negative emotions, low FL scores, and high-risk customers.

A novel Self-Organizing Neuro-Fuzzy Multilayered Classifier, the GA-SONeFMUC model, is proposed in this paper for land cover classification of multispectral images. The model is composed of generic fuzzy neuron classifiers (FNCs) arranged in layers, which are implemented by fuzzy rule-based systems. At each layer, parent FNCs are combined to generate a descendant FNC at the next layer with higher classification accuracy. To exploit the information acquired by the parent FNCs, their decision supports are combined using a fusion operator. As a result, a data splitting is devised within each FNC, distinguishing those pixels that are currently correctly classified to a high certainty grade from the ambiguous ones. The former are handled by the fuser, while the ambiguous pixels are further processed to enhance their classification confidence. The GA-SONeFMUC structure is determined in a self-constructing way via a structure-learning algorithm with feature selection capabilities. The parameters of the models obtained after structure learning are optimized using a real-coded genetic algorithm. For effective classification, we formulated three input sets containing spectral and textural feature types. To explore information coming from different feature sources, we apply a classifier fusion approach at the final stage. The outputs of individual classifiers constructed from each input set are combined to provide the final assignments. Our approach is tested on a lake-wetland ecosystem of international importance using an IKONOS image. A high-classification performance of 92.02% and of 75.55% for the wetland zone and the surrounding agricultural zone is achieved, respectively.

In this paper, we defined the viseme (visual speech element) and described about the method of extracting visual feature vector. We defined the 10 visemes based on vowel by analyzing of Korean utterance and proposed the method of extracting the 20-dimensional visual feature vector, combination of static features and dynamic features. Lastly, we took an experiment in recognizing words based on 3-viseme HMM and evaluated the efficiency.

This paper investigates channel-aware decision fusion empowered
by massive MIMO systems and reconfigurable intelligent surfaces
(RIS). By integrating both, we aim to improve goal-oriented (fusion)
performance despite the unique propagation challenges introduced.
Specifically, we investigate traditional favorable propagation properties
in the context of RIS-aided Massive MIMO decision fusion. The
above analysis is then leveraged (i) to design three sub-optimal simple
fusion rules suited for the large-array regime and (ii) to devise an
optimization criterion for RIS reflection coefficients based on longterm
channel statistics. Simulation results confirm the appeal of the
presented design.

Uncertainty representation is a major issue in pattern recognition. In many applications, the outputs of a classifier do not lead directly to a final decision, but are used in combination with other systems, or as input to an interactive decision process. In such contexts, it may be advantageous to resort to rich and flexible formalisms for representing and manipulating uncertain information. This paper addresses the issue of uncertainty representation in pattern classification, in the framework of the Dempster-Shafer theory of evidence. It is shown that the quality and reliability of the outputs of a classifier may be improved using a variant of bagging, a resample-and-combine approach introduced by Breiman in a conventional statistical context. This technique is explained and studied experimentally on simulated data and on a character recognition application. In particular, results show that bagging improves classification accuracy and limits the influence of outliers and ambiguous training patterns.

The detector uses phase compensation for detecting sinusoidal signal with unknown amplitude, frequency, and phase in complex Gaussian noise with unknown variance. We analyze the statistical properties of the discrete Fourier transform (DFT) of the observations and consider the leakage effect. Then we develop the detector by using generalized likelihood ratio test (GLRT). The new detector fully utilizes the phase information of the signal and achieves constant false alarm rate (CFAR) property. Theoretical analysis and simulation results show its improvement in detection performance.

The aim of this paper is to present a method for source detection within unknown chemical mixtures using several measurement modalities. Contrary to the well studied case of single source detection, this approach enables simultaneous detection of multiple chemical components by exploiting the mixing coefficients resulting from supervised linear unmixing and thresholded non-negative least squares. The first contribution of this work is to propose an automated procedure to compute an optimized binary classifier rule for each component independently using a database of known mixtures. The second contribution is to propose a global decision rule based on the fusion of the multimodal decisions using weighting schemes such as those used in multiple classifier systems (MCS). A real database of Ion Mobiliy Mass Spectrometry (IMMS) data is used to evaluate the detection performance. An increase of the detection accuracy is reached using the multiple thresholds within the independent classifiers approach as compared to single modality detection.

Dierent learning algorithms for the decision fusion mapping of a multiple classier system are compared in this paper. It is very well known that the confusion matrices of the individual classiers are utilised in the naive Bayes combination of classier outputs. After a brief review of the decision templates, the linear associative memory and the pseudoinverse matrix approaches it is demonstrated that all four adaptive decision fusion mappings share the confusion matrices as the essential ingredient.

We consider the problem of robust detection of a spread-spectrum (SS) signal in the presence of unknown correlated SS interference and additive non-Gaussian noise. The proposed general SS receiver structure is comprised by a vector of adaptive chip-based nonlinearities followed by an adaptive linear tap-weight filter and combines the relative merits of both nonlinear and linear signal processing. The novel characteristics of our approach are as follows. First, the nonlinear receiver frontend adapts itself to the unknown prevailing noise environment providing robust performance for a wide range of underlying noise distributions. Second, the adaptive linear tap-weight filter that follows the nonlinearly processed chip samples results in a receiver that is proven to be effective in combating SS interference as well. To determine the receiver parameters, we propose, develop, and study three adaptive schemes under a joint meansquare-error (MSE), or a joint bit-error-rate (BER), or a joint MSE-BER optimization criterion. As a side result, we derive the optimum decision fusion filter for receivers that utilize hardlimiting (sign) chip nonlinearities. Numerical and simulation results demonstrate the performance of the proposed schemes and offer comparisons with the conventional matched-filter (MF), the decorrelator, the conventional minimum-variance-distortionlessresponse (MVDR) filter, and the sign-majority-vote receiver.

Burn injuries present a substantial global public health challenge. The conventional approach, relying on visual inspection to compute total body surface area (TBSA) for assessing burn severity, encounters the inherent limitations of proper estimation. These limitations prompted the development of computer-based applications, particularly machine learning, and deep learning models, to enhance the performance. This paper presents a comprehensive analytical study of eight deep learning techniques designed for assessing burn severity in terms of four characteristics: inflammation, scar, uniformity, and pigmentation, in small datasets of 2-dimensional (2D) images captured using digital (smartphone) camera. The models are Convolutional Neural Network (CNN), attention-based CNN, decision-level fusion (DF) based on CNN models, DF with attention-based CNN models, autoencoder-NN (Neural Network), and hybrid VGG16-Machine Learning (ML) with Support Vector Machine (SVM), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost). Each model is validated with eight datasets collected and annotated by our team at the Children’s Hospital of Michigan in two phases to classify the severity of burns in terms of inflammation, scar, uniformity, and pigmentation. The average test accuracy across the eight datasets using CNN, attention-based CNN, DF with CNN models, DF with attention-based CNN models, autoencoder-NN, VGG16-RF, VGG16-SVM, and VGG16-XGBoost are $0.87\pm 0.04$ , $0.93\pm 0.04$ , $0.90\pm 0.01$ , $0.95\pm 0.02$ , $0.87\pm 0.03$ , $0.63\pm 0.03$ , $0.79\pm 0.02$ , $0.79\pm 0.01$ , correspondingly. The research also computes and compares the computational complexity of each model in terms of FLoating point Operations Per Second (FLOPS) and Multiply-ACcumulate operations (MACs). Compared with the base CNN model, the decision-level fusion with attention mechanism model outperforms with a gain of 9.19% in test accuracy and an increase of 3321.53% in FLOPS. Considering the priority and constraint of the task, the attention-based CNN model can also be preferable as it achieves an accuracy gain of 6.90% and significantly less computational increase expense (8.62%) compared with the base CNN. The code for the best performing decision-level fusion with attention mechanism model is provided on GitHub link at https://github.com/Saeka2022/Burn-Assessment .

Nonlinear mapping has been used in the past for data structure analysis. Many interesting heuristic approaches to map n-dimensional data to a lower dimensional space such that the local structure of the original data is reserved have been proposed by Kruskal, Sammon, and others. Basically, a criterion for local structure preservation is first defined, then a new data configuration is obtained by an iterative process to minimize the selected criterion.

Intrusion Detection System (IDS) technology is an important component in designing a secure environment. Alert aggregation is an important subtask of intrusion detection. The goal is to identify and to cluster different alerts produced by lowlevel intrusion detection systems, firewalls, etc. belonging to a specific attack instance which has been initiated by an attacker at a certain point in time. Thus, meta-alerts can be generated for the clusters that contain all the relevant information whereas the amount of data (i.e., alerts) can be reduced substantially. Distributed IDS systems are the next logical level for IDS systems to move to. A distributed IDS (dIDS) consists of multiple Intrusion Detection Systems (IDS) over a large network, all of which communicate with each other, or with a central server that facilitates advanced network monitoring, incident analysis, and instant attack data. A dIDS also allows to identify threats to the network across multiple network segments. In N...

We consider decentralized detection of an unknown signal corrupted by zero-mean unimodal noise via wireless sensor networks. We assume the presence of both smart and dumb sensors: the former transmit unquantized measurements, whereas the latter employ multilevel quantizations (before transmission through binary symmetric channels) in order to cope with energy and/or bandwidth constraints. The data are received by a fusion center, which relies on a proposed Rao test, as a simpler alternative to the generalized likelihood ratio test (GLRT). The asymptotic performance analysis of the multibit Rao test is provided and exploited to propose a (signal-independent) quantizer design approach by maximizing the noncentrality parameter of the test-statistic distribution. Since the latter is a nonlinear and nonconvex function of the quantization thresholds, we employ the particle swarm optimization algorithm for its maximization. Numerical results are provided to show the effectiveness of the Rao test in comparison to the GLRT and the boost in performance obtained by (multiple) threshold optimization. Asymptotic performance is also exploited to define detection gain measures allowing to assess gain arising from use of dumb sensors and increasing their quantization resolution.

We consider a decentralized multi-sensor estimation problem where sensor nodes observe noisy versions of a correlated random source vector. The sensors amplify and forward their observations over a fading coherent multiple access channel (MAC) to a fusion center (FC). The FC is equipped with a large array of antennas and adopts a minimum mean-square error (MMSE) approach for estimating the source. We optimize the amplification factor (or equivalently transmission power) at each sensor node in two different scenarios: a) with the objective of total power minimization subject to mean square error (MSE) of source estimation constraint, and b) with the objective of minimizing MSE subject to total power constraint. For this purpose, based on the well-known favorable propagation condition (when) achieved in massive multiple-input multiple-output (MIMO), we apply an asymptotic approximation on the MSE and use convex optimization techniques to solve for the optimal sensor power allocation in a) and b). In a), we show that the total power consumption at the sensors decays as , replicating the power savings obtained in massive MIMO mobile communications literature. We also show several extensions of the aforementioned scenarios to the cases where sensor-to-FC fading channels are correlated, and channel coefficients are subject to estimation error. Through numerical studies, we also illustrate the superiority of the proposed optimal power allocation methods over uniform power allocation.

Multichannel analysis is an innovative technique used for the analysis of bioelectrical signals. In this paper, we analyzed uterine Electromyogram (EMG) signals recorded by means of a 4x4 electrode matrix positioned on the woman's abdomen by using a multichannel approach. Relevant features were extracted from each channel and fed to a competitive neural network (CNN). First, we evaluated the classification performance of each channel. Then, we compared these performances to see which channel ranks better than the others. Finally, a decision fusion method based on the weighted sum of the individual decision of each channel was tested. The results showed that data can be grouped into 2 different groups. Furthermore, they showed that the classification performance varies according to the position of the electrode. Therefore, when a decision fusion rule was applied, the network yielded better classification accuracy than any individual channel could provide. These encouraging results prove that multichannel analysis can improve the classification of uterine EMG signals.