Abstract The Canary Islands, internationally recognized as a biodiversity hot spot, possess unique ecological characteristics including endemic reptile species that face substantial threats from invasive alien species. Particularly, concerning is the California kingsnake ( Lampropeltis californiae ), which exhibits remarkable adaptability and inflicts severe ecological damage on endangered endemic fauna. This proposal aims to address biodiversity conservation challenges in the Canary Islands ecosystem by advancing reptile species classification through the implementation of Swin Transformer architectures. The integration of these advanced neural network architectures with conservation biology provides an automated, precise tool for species identification across different taxonomic levels that can enhance monitoring and control strategies for biodiversity preservation worldwide. This technical approach addresses urgent conservation requirements in the Canary Islands while simultaneously establishing a methodological framework applicable to other ecological contexts facing comparable biodiversity threats. The methodology employed pre-trained Swin Transformer models, initially developing fine-grained reptile species discrimination capabilities through hyperparameter optimization on a limited-scale database (160 images, 40 per species) using nested cross-validation to ensure statistical rigor and independence. The optimized configuration was subsequently transferred to a substantially larger database (4,400 images) for binary classification distinguishing snake from non-snake reptiles, where training on only 400 samples achieved 99.38% accuracy on 4,000 independent test samples, directly addressing the invasive alien species monitoring challenge. The Swin-B Transformer model demonstrated robust performance, achieving up to 100% accuracy in multiclass species classification and 99.38% accuracy in the challenging, taxonomically broader binary snake versus non-snake classification. These results establish the transferability and scalability of the hyperparameter optimization methodology across datasets of substantially different magnitudes.

Deep Learning methods require immense amounts of labeled data to provide reasonable results. In computer vision applications, and more specifically in despeckling SAR (Synthetic Aperture Radar) images, due to the speckle content, there is no ground truth available. To test the performances of despeckling filters, the common protocol is to synthetically corrupt optical images with a suitable speckle model and then, after filtering, well-known metrics are obtained. Then, filters are tested on actual SAR data. However, even the most elaborated speckle models are far from accounting for the complex mechanisms related to SAR images. In this paper, a methodology to design a realistic dataset is proposed. Actual SAR images of the same scene acquired with the same sensor on different dates, then they are properly co-registered and averaged to get a ground truth-like reference image to objectively evaluate the performance of a despeckling method. To show the benefits of the proposed methodology, a deep learning approach is used to filter the data by using the designed dataset, which will be called the “SAR model”. Then they are compared with the standard protocol by using synthetically corrupted optical images, which will be the “Synthetic model”. One last validation is performed by filtering the same images with FANS, a well-known despeckling filter and compared with the results obtained with autoencoder. The validation on actual SAR data not included in the training phase validates the proposed methodology. From the results shown, it is recommended to test filters on the proposed more realistic dataset.

We present a novel approach of 2D to 3D transfer learning based on mapping pre-trained 2D convolutional neural network weights into planar 3D kernels. The method is validated by the proposed planar 3D res-u-net network with encoder transferred from the 2D VGG-16, which is applied for a single-stage unbalanced 3D image data segmentation. In particular, we evaluate the method on the MICCAI 2016 MS lesion segmentation challenge dataset utilizing solely fluid-attenuated inversion recovery (FLAIR) sequence without brain extraction for training and inference to simulate real medical praxis. The planar 3D res-u-net network performed the best both in sensitivity and Dice score amongst end to end methods processing raw MRI scans and achieved comparable Dice score to a state-of-the-art unimodal not end to end approach. Complete source code was released under the open-source license, and this paper complies with the Machine learning reproducibility checklist. By implementing practical transfer learning for 3D data representation, we could segment heavily unbalanced data without selective sampling and achieved more reliable results using less training data in a single modality. From a medical perspective, the unimodal approach gives an advantage in real praxis as it does not require co-registration nor additional scanning time during an examination. Although modern medical imaging methods capture high-resolution 3D anatomy scans suitable for computer-aided detection system processing, deployment of automatic systems for interpretation of radiology imaging is still rather theoretical in many medical areas. Our work aims to bridge the gap by offering a solution for partial research questions.

The objective of this paper is to present the state of-the-art relating to automatic speech and voice analysis techniques as applied to the monitoring of patients suffering from Alzheimer's disease as well as to shed light on possible future research topics. This work reviews more than 90 papers in the existing literature and focuses on the main feature extraction techniques and classification methods used. In order to guide researchers interested in working in this area, the most frequently used data repositories are also given. Likewise, it identifies the most clinically relevant results and the current lines developed in the field. Automatic speech analysis, within the Health 4.0 framework, offers the possibility of assessing these patients, without the need for a specific infrastructure, by means of non-invasive, fast and inexpensive techniques as a complement to the current diagnostic methods.

Genetic and neurophysiological studies of electroencephalogram (EEG) have shown that an individual's brain activity during a given cognitive task is, to some extent, determined by their genes. In fact, the field of biometrics has successfully used this property to build systems capable of identifying users from their neural activity. These studies have always been carried out in isolated conditions, such as relaxing with eyes closed, identifying visual targets or solving mathematical operations. Here we show for the first time that the neural signature extracted from the spectral shape of the EEG is to a large extent independent of the recorded cognitive task and experimental condition. In addition, we propose to use this task-independent neural signature for more precise biometric identity verification. We present two systems: one based on real cepstrums and one based on linear predictive coefficients. We obtained verification accuracies above 89% on 4 of the 6 databases used. We anticipate this finding will create a new set of experimental possibilities within many brain research fields, such as the study of neuroplasticity, neurodegenerative diseases and brain machine interfaces, as well as the mentioned genetic, neurophysiological and biometric studies. Furthermore, the proposed biometric approach represents an important advance towards real world deployments of this new technology.

The recent evolution of storage devices, digital embedded cameras and the Internet have collaterally allowed sexual predators to take advantage of these technological breakthroughs to gather illegal media, which is exhibited uncensored through Peer-to-Peer file sharing networks. In this paper, we are particularly concerned about the increasing availability of Child Abuse Material. Therefore, we have explored alternatives to detect non-adults in visual content. Initially, different age estimations and underage detection techniques are reviewed by analyzing existing datasets. Finally, several local descriptors and Convolutional Neural Networks for underage detection are evaluated. The experimental results obtained for a large dataset that combines collections such as FG-Net, Adience, GenderChildren, The Image of Groups and Boys2Men evidence the complementary information contained in both local descriptors and neural networks , as their fusion boosts the accuracy of non-adult detection to over 93%.

Nowadays, there is an increasing interest in wireless sensor networks (WSN) for environmental monitoring systems because it can be used to improve the quality of life and living conditions are becoming a major concern to people. This paper describes the design and development of a real time monitoring system based on ZigBee WSN characterized by a lower energy consumption, low cost, reduced dimensions and fast adaptation to the network tree topology. The developed system encompasses an optimized sensing process about environmental parameters, low rate transmission from sensor nodes to the gateway, packet parsing and data storing in a remote database and real time visualization through a web server. A monitoring system integrating the outlined system has been deployed and tested for monitoring the level of dust particles in the air, acoustic levels in different places of a city, ambient temperature and relative humidity. A calibration process of a low cost audio sensor was performed to measure the acoustic level from different noise sources, hence, it is not necessary to use an expensive sound level meter at each node. Furthermore, experimental results show autonomy nodes can be about three months.

The objective of this paper is to provide a review of recent advances in automatic vibration- and audio-based fault diagnosis in machinery using condition monitoring strategies. It presents the most valuable techniques and results in this field and highlights the most profitable directions of research to present. Automatic fault diagnosis systems provide greater security in surveillance of strategic infrastructures, such as electrical substations and industrial scenarios, reduce downtime of machines, decrease maintenance costs, and avoid accidents which may have devastating consequences. Automatic fault diagnosis systems include signal acquisition, signal processing, decision support, and fault diagnosis. The paper includes a comprehensive bibliography of more than 100 selected references which can be used by researchers working in this field.

A method for Off-line Handwritten Signature Verification is described. It works at the global and local image level, measuring the stroke gray-level variations by means of wavelet analysys and statistical texture features. This method begins with a proposed background removal. Then Wavelet Analysis allows to estimate and alleviate the global influence of ink-type, and finally, properties of the Co-occurrence Matrix are used as features representing individual characteristics at local level. Genuine samples have been used for train an SVM model, random and skilled forgeries have been used for testing it. Experiments were conducted on three differente databases (MCYT75, GPDS100, and GPDS750). Results are reasonable according to the state of the art and approaches that use the same public available database (MCYT75) and prove the feasibility of the proposed methodology.

Features representing information about pressure distribution from a static image of a handwritten signature are analyzed for an offline verification system. From gray-scale images, its histogram is calculated and used as "spectrum'' for calculation of pseudo-cepstral coefficients. Finally, the unique minimum-phase sequence is estimated and used as feature vector for signature verification. The optimal number of pseudo-coefficients is estimated for best system performance. Experiments were carried out using a database containing signatures from 100 individuals. The robustness of the analyzed system for simple forgeries is tested out with a LS-SVM model. For the sake of completeness, a comparison of the results obtained by the proposed approach with similar works published using pseudo-dynamic feature for offline signature verification is presented.

In this paper, we propose to quantify the quality of the recorded voice through objective nonlinear measures. Quantification of speech signal quality has been traditionally carried out with linear techniques since the classical model of voice production is a linear approximation. Nevertheless, nonlinear behaviors in the voice production process have been shown. This paper studies the usefulness of six nonlinear chaotic measures based on nonlinear dynamics theory in the discrimination between two levels of voice quality: healthy and pathological. The studied measures are first- and second-order Renyi entropies, the correlation entropy and the correlation dimension. These measures were obtained from the speech signal in the phase-space domain. The values of the first minimum of mutual information function and Shannon entropy were also studied. Two databases were used to assess the usefulness of the measures: a multiquality database composed of four levels of voice quality (healthy voice and three levels of pathological voice); and a commercial database (MEEI Voice Disorders) composed of two levels of voice quality (healthy and pathological voices). A classifier based on standard neural networks was implemented in order to evaluate the measures proposed. Global success rates of 82.47% (multiquality database) and 99.69% (commercial database) were obtained.

The current need for large databases to evaluate automatic biometric recognition systems has motivated the developing of the GPDS-960 corpus, an off-line handwritten signature database which contains 24 genuine signatures and 30 forgeries of 960 individuals. This paper describes the GPDS signature corpus, gives details about the acquisition protocols and presents preliminary verification results obtained using the GPDS data.

The effect of changing the image resolution over an off-line signature verification system performance is analyzed. The geometrical features used for the system analyzed in this paper are based on two vectors which represent the envelope and the interior stroke distribution in polar and Cartesian coordinates. Image resolution is progressively diminished from an initial 600 ppp resolution till 45 ppp. The robustness of the analyzed system for random and simple forgeries is tested out with a hidden Markov model. The results show that 150 ppp offers a good trade-off between performance and image resolution for static features.

This paper presents a set of geometric signature features for offline automatic signature verification based on the description of the signature envelope and the interior stroke distribution in polar and Cartesian coordinates. The features have been calculated using 16 bits fixed-point arithmetic and tested with different classifiers, such as hidden Markov models, support vector machines, and Euclidean distance classifier. The experiments have shown promising results in the task of discriminating random and simple forgeries.