A5011988914- Human Pose and Action Recognition
- Advanced Image and Video Retrieval Techniques
- Video Analysis and Summarization
- Gait Recognition and Analysis
- Hand Gesture Recognition Systems
- Video Surveillance and Tracking Methods
- Anomaly Detection Techniques and Applications
- Face recognition and analysis
- Music and Audio Processing
- Image Retrieval and Classification Techniques
- Biometric Identification and Security
- Medical Image Segmentation Techniques
- Online Learning and Analytics
- Multimodal Machine Learning Applications
- Handwritten Text Recognition Techniques
- Network Security and Intrusion Detection
- Hearing Impairment and Communication
- Generative Adversarial Networks and Image Synthesis
- Visual Attention and Saliency Detection
- AI in cancer detection
- Algorithms and Data Compression
- Advanced Image Processing Techniques
- Vehicle License Plate Recognition
- Advanced Neural Network Applications
- Reconstructive Facial Surgery Techniques
Universidade Federal de Ouro Preto
2014-2023
Universidad Católica San Pablo
2022
Universidade Federal de Minas Gerais
2006-2010
École Nationale Supérieure de l'Électronique et de ses Applications
2006-2008
Centre National de la Recherche Scientifique
2008
Universidade Presidente Antônio Carlos
2006
Icesi University
2004
An electrocardiogram (ECG) measures the electric activity of heart and has been widely used for detecting diseases due to its simplicity non-invasive nature. By analyzing electrical signal each heartbeat, i.e., combination action impulse waveforms produced by different specialized cardiac tissues found in heart, it is possible detect some abnormalities. In last decades, several works were developed produce automatic ECG-based heartbeat classification methods. this work, we survey current...
Deaf people use systems of communication based on sign language and finger spelling. Manual spelling, or is a system where each letter the alphabet represented by an unique discrete movement hand. RGB depth images can be used to characterize hand shapes corresponding letters alphabet. The advantage cameras over color for gesture recognition more evident when performing segmentation. In this paper, we propose hybrid approach spelling using RGB-D information from Kinect sensor. first stage,...
Automatic violence detection in video surveillance is essential for social and personal security. Monitoring the large number of cameras used public private areas challenging human operators. The manual nature this task significantly increases possibility ignoring important events due to limitations when paying attention multiple targets at a time. Researchers have proposed several methods detect violent automatically overcome problem. So far, most previous studies focused only on...
The continuous creation of digital video has caused an exponential growth content. To increase the usability such large volume videos, a lot research been made. Video summarization proposed to rapidly browse collections. summarize any type video, researchers have relied on visual features contained in frames. In order extract these features, different techniques used local or global descriptors. this paper, we propose method for static that can produce meaningful and informative summaries....
In this paper, we propose a new approach for dynamic hand gesture recognition using intensity, depth and skeleton joint data captured by Kinect <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TM</sup> sensor. The proposed integrates global local information of gesture. First, represent the 3D trajectory in spherical coordinates. Then, extract key frames corresponding to points with more angular distance difference. each frame, calculate from...
Automatic violence detection in video surveillance is crucial for social and personal security. Due to the massive data produced by cameras installed different environments like airports, trains, stadiums, schools, etc., traditional monitoring humans operators becomes inefficient. In this context, develop systems capable of detect automatically violent actions a challenging task. This study describes method localize acts using dynamic images, CNN's, weakly supervised localization methods....
The prediction of university dropout is a complex problem, given the number and diversity variables involved. Therefore, different strategies are applied to understand this educational phenomenon, although most outstanding derive from joint application statistical approaches computational techniques based on machine learning. Student Dropout Prediction (SDP) challenging problem that can be addressed following various strategies. On one hand, learning formulate it as classification task whose...
In this paper, we present a new approach for dynamic hand gesture recognition that uses intensity, depth, and skeleton joint data captured by Kinect sensor. This method integrates global local information of gesture. First, represent the 3D trajectory in spherical coordinates. Then, select most relevant points with our proposed keyframe detection. After, movements spatial, temporal position changes information. Next, use direction cosines definition to describe body positions generating...
Recognition using ear images has been an active field of research in recent years. Besides faces and fingerprints, ears have a unique structure to identify people can be captured from distance, contactless, without the subject's cooperation. Therefore, it represents appealing choice for building surveillance, forensic, security applications. However, many techniques used those applications-e.g., convolutional neural networks (CNN)-usually demand large-scale datasets training. This work...
Computer science is a dynamic field of study that requires constant review and updating the curricular designs in academic programs—in general, measuring impact plan changes has been little explored literature. In most cases, it focuses only on structuring its curricula, leaving aside several factors associated with important events or facts such as student dropout, retention, inclusion. However, these features provide institutions many opportunities to understand performance propose more...
Due to the increasingly need for automatic traffic monitoring, vehicle license plate detection is of high interest perform toll collection, law enforcement, parking lot access control, among others. In this paper, a sliding window approach based on Histogram Oriented Gradients (HOG) features used Brazilian detection. This consists in scanning whole image multiscale fashion such that located precisely. The main contribution work deep study best setup HOG descriptors plates, which have never...
Video shot boundary detection plays an important role in video processing. It is the first step toward video-content analysis and content-based retrieval. We develop a hierarchical approach for based on assumption that hierarchy helps to take decisions by reducing amount of indeterminate transitions. Our method consists detecting abrupt transitions using learning-based approach, then non-abrupt are split into gradual normal frames. describe this paper, machine learning system detection. The...
Sign language recognition (SLR) is a very challenging task due to the complexity of learning or developing descriptors represent its primary parameters (location, movement, and hand configuration). In this paper, we propose robust deep based method for sign recognition. Our approach represents multimodal information (RGB-D) through texture maps describe location movement. Moreover, introduce an intuitive extract representative frame that describes shape. Next, use as inputs two three-stream...
The mosquito Aedes aegypti can transmit some diseases, which makes the study of proliferation this vector a necessary task. With use traps made in laboratory, called ovitraps, it is possible to map egg deposition community. Through camera, coupled with magnifying glass, are acquired images containing elements (eggs) be counted. First, goal find pixels similar color eggs; for that, we take advantage slice method. From these already worked images, process transfer learning convolutional neural...
High and persistent dropout rates represent one of the biggest challenges for improving efficiency educational system, particularly in underdeveloped countries. A range features influence college dropouts, with some belonging to field others non-educational fields. Understanding interplay these variables identify a student as potential could help decision makers interpret situation decide what they should do next reduce based on corrective actions. This paper presents SDA-Vis, visualization...
The paper provides a summary of the 2023 Unconstrained Ear Recognition Challenge (UERC), benchmarking effort focused on ear recognition from images acquired in uncontrolled environments. objective challenge was to evaluate effectiveness current techniques challenging dataset while analyzing two distinct aspects, i.e., verification performance and bias with respect specific demographic factors, gender ethnicity. Seven research groups participated submitted seven approaches that ranged...
Systems for face sketch recognition are very important law enforcement agencies. These systems can help to locate or narrow down potential suspects. Recently, various methods were proposed address this problem, but there is no clear comparison of their performance. In paper a new approach photo/sketch based on the Local Feature-based Discriminant Analysis (LFDA) method. This was tested and compared with its predecessors using three differents datasets also adding an extra gallery 10,000...
In this paper, we propose a new approach for finger spelling recognition using depth information captured by Kinect sensor. We only use to characterize hand configurations corresponding alphabet letters. First, data generate binary mask which is used segment the area from background. Then, major axis determined and aligned with Y in order achieve rotation invariance. Later, convert 3D point cloud. The cloud divided into sub regions each one, direction cosines, calculated three histograms of...
Deaf people use systems of communication based on sign language and finger spelling. Finger spelling is a system where each letter the alphabet represented by unique discrete movement hand. RGB depth images can be used to characterize hand shapes corresponding letters alphabet. There exists an advantage sensors, as Kinect, over color cameras for recognition: provide 3D information In this paper, we propose model recognition using kernel descriptors, consisting four stages. The performance...
The recent progress of digital media has stimulated the creation, storage and distribution data, such as videos, generating a large volume data requiring efficient technolo- gies to increase usability these data. Video summarization methods generate concise summaries video contents enable faster browsing, indexing accessing collections, however, often perform slow with duration high quality One way reduce this long time execution is develop parallel algorithm, using advantages computer...