A5013906184- Biometric Identification and Security
- Image Retrieval and Classification Techniques
- Face and Expression Recognition
- ECG Monitoring and Analysis
- Advanced Data Compression Techniques
- Natural Language Processing Techniques
- Face recognition and analysis
- Authorship Attribution and Profiling
- AI in cancer detection
- Algorithms and Data Compression
- Digital Media Forensic Detection
- User Authentication and Security Systems
- Forensic Fingerprint Detection Methods
- Non-Invasive Vital Sign Monitoring
- EEG and Brain-Computer Interfaces
- Analog and Mixed-Signal Circuit Design
- Advanced Image and Video Retrieval Techniques
- Topic Modeling
- Fire Detection and Safety Systems
- Flood Risk Assessment and Management
- Image Processing and 3D Reconstruction
- Water Quality and Pollution Assessment
- Medical Image Segmentation Techniques
- Inertial Sensor and Navigation
- Image and Video Quality Assessment
Universidade Federal da Paraíba
2015-2024
CETESB - Companhia Ambiental do Estado de São Paulo
2023
Brazilian Institute of Environment and Renewable Natural Resources
2023
Águas de Portugal (Portugal)
2023
Universidade Federal do Pará
2022
Universidade Federal de Campina Grande
2018
Genomic Islands (GIs) are regions of bacterial genomes that acquired from other organisms by the phenomenon horizontal transfer. These often responsible for many important adaptations bacteria, with great impact on their evolution and behavior. Nevertheless, these usually associated pathogenicity, antibiotic resistance, degradation metabolism. Identification such is medical industrial interest. For this reason, different approaches genomic islands prediction have been proposed. However, none...
This paper presents a face recognition method based on Discrete Cosine Transform (DCT) coefficient's selection. Without normalization phase, the proposed uses, in its feature selection stage, technique only DCT coefficients amplitudes. Three coefficient criterions were analyzed: first one is average of coefficients' amplitudes; second counting occurrence each coefficient, which are stored set lists containing most significant coefficients; finally, third criterion position list ordered by...
Automatic face recognition is a challenging problem, since human faces have complex pattern. This paper presents technique for of frontal on gray scale images. In this technique, the distance between Discrete Cosine Transform (DCT) under evaluation and all DCTs database are computed. The with shortest distances probably belong to same person; therefore evaluating attributed person. calculated as sum differences modules DCT coefficients. Only few coefficients used in computation; they...
A general-purpose unsupervised segmentation algorithm based on cross-entropy minimization by pixel was developed; this algorithm, known as the SCEMA (Segmentation Cross-Entropy Minimization Algorithm), starts from an initial and iteratively searches best statistical model, estimating probability density of image to reduce with respect previous iteration. The tested using satellite images Landsat Thematic Mapper sensor 5 for Amazon region (12 testing 15 validation). Theme classes identified...
The collapse of Dam I, owned by Vale S.A, in Brumadinho-MG (Brazil), among other serious socio-environmental consequences, contaminated the waters Paraopeba River a stretch hundreds kilometers. Considering relevance monitoring water quality, and knowing that field evaluation is time-consuming costly procedure, use satellite images, widely available at low cost, emerges as relevant alternative. This work proposes systematic experimental five machine learning methods - Extra Trees, Multilayer...
In recent years, the demand for facial biometric authentication services has increased dramatically. Also, efforts to cheat this type of system have become more common. paper, we propose a single shot CNN-based solution face anti-spoofing problem. We trained deep learning model using transfer from pre-trained VGG16 model. After some pre-processing rely solely on network classify an image. evaluate several implications preprocessing data, investigate different amounts background included in...
In this work we discuss author identification for documents written in Portuguese. Two different approaches were compared. The first is the writer-independent model which reduces pattern recognition problem to a single and two classes, hence, makes it possible build robust system even when few genuine samples per writer are available. second personal model, very often performs better but needs bigger number of writer. We also introduce stylometric feature set based on conjunctions adverbs...
The aim of this study was to histologically assess the effect laser therapy (LILT, 660 and 780 nm) on repair standardized bone defects femur Wistar albinus rats. sample composed 12 young adult rats both genders. Three randomized groups were studied: group I (control, n = 4), II nm, III 4). Samples prepared using a defect left-side surface animals, with total dimension approximately 3 mm3. Groups irradiated every 48 h from second application, where first dose given immediately after surgery...
This paper proposes an ECG signal compressor based on optimum quantization of discrete cosine transform (DCT) coefficients and Golomb-Rice coding. The to be compressed is initially partitioned in blocks, each DCT block quantized using a step size vector zeroing threshold vector. These vectors are defined so that the estimated entropy minimized for target distortion reconstructed or, alternatively, entropy. final To assess performance compressor, records MIT-BIH Arrhythmia Database were at...
In this paper we discuss the use of compression algorithms for author identification. We present basic background about and introduce prediction by partial matching algorithm, which has been used in our experiments. To better compare results produced PPM some experiments using stylometric features very often forensic examiners. case authors are modeled support vector machines. Comprehensive performed on a database composed 20 different show that algorithm is an interesting alternative...
Seismic image interpretation is indispensable for oil and gas industry. Currently, artificial intelligence has been undertaken to increase the level of confidence in exploratory activities. Detecting potentially recoverable hydrocarbon zones (leads) under viewpoint computer vision an emerging problem that demands thorough examination. This paper introduces a processing workflow recognize geologic leads seismic images resorts encoder-decoder architectures convolutional neural network (CNN)...
This work presents a new multiscale, curvature-based shape representation technique for planar curves. One limitation of the well-known curvature scale space (CSS) method is that it uses only zero-crossings to characterize shapes and thus there no CSS descriptor convex shapes. The proposed method, on other hand, bidimensional-unidimensional-bidimensional transformations together with resampling techniques retain full information characterization. It also employs correlation coefficient as...
In this paper we compare two different paradigms for author identification. The first one is based on compression algorithms where the entire process of defining and extracting features training a classifier avoided. second paradigm, other hand, takes into account classical pattern recognition framework, linguistic proposed by forensic experts are used to train Support Vector Machine classifier. Comprehensive experiments performed database composed 20 writers show that both strategies...
This paper presents an approach for personal recognition based on hand geometry applying different classification and training methods to measure the results. The features extraction process prioritizes user comfort during capture produces segmentation of hands fingers with high precision. For classification, Bayesian networks support vector machines were applied in three implementations. Tests using cross-validation random subsampling techniques performed. experiments demonstrated...
This paper describes a method of heart arrhythmia classification based on the rate variability (HRV) signal and compression algorithm Prediction by Partial Matching. The arrhythmias to be identified are: Normal Sinus Rhythm, Premature Ventricular Contraction, 2 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">nd</sup> Heart Block Bradycardia. extraction HRV is performed analyzing electrocardiogram detect R peak from QRS complex heartbeats, then...
Abstract Background Quantification of tumor heterogeneity is essential to better understand cancer progression and adapt therapeutic treatments patient specificities. Bioinformatic tools assess the different cell populations from single-omic datasets as bulk transcriptome or methylome samples have been recently developed, including reference-based reference-free methods. Improved methods using multi-omic are yet be developed in future community would need systematic perform a comparative...