A5078836003- Music and Audio Processing
- Handwritten Text Recognition Techniques
- Music Technology and Sound Studies
- Biometric Identification and Security
- Hand Gesture Recognition Systems
- Forensic Fingerprint Detection Methods
- Speech and Audio Processing
- Human Pose and Action Recognition
- Hearing Impairment and Communication
- Gait Recognition and Analysis
- Industrial Vision Systems and Defect Detection
- Image Processing and 3D Reconstruction
- Face recognition and analysis
- Emotion and Mood Recognition
- User Authentication and Security Systems
- Face and Expression Recognition
- Image and Video Stabilization
- EEG and Brain-Computer Interfaces
- Image Retrieval and Classification Techniques
- Remote-Sensing Image Classification
- Automated Road and Building Extraction
- Neural Networks and Applications
- Tactile and Sensory Interactions
- Sleep and Work-Related Fatigue
- 3D Shape Modeling and Analysis
Institute for Systems Engineering and Computers
2011-2023
INESC TEC
2008-2022
Universidade Portucalense
2018-2020
Universidade do Porto
2008-2019
Aquatic Systems (United States)
2014
Innovation Performance (Norway)
2014
Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento
2007
The preservation of musical works produced in the past requires their digitalization and transformation into a machine-readable format. processing handwritten scores by computers remains far from ideal. One fundamental stages to carry out this task is staff line detection. We investigate general-purpose, knowledge-free method for automatic detection music lines based on stable path approach. Lines affected curvature, discontinuities, inclination are robustly detected. Experimental results...
The importance of recycling is well known, either for environmental or economic reasons, it impossible to escape and the industry demands efficiency. Manual labour traditional industrial sorting techniques are not capable keeping up with objectives demanded by international community. Solutions based in computer vision have potential automate part waste handling tasks. In this paper, we propose a hierarchical deep learning approach detection classification food trays. proposed two-step...
Biometrics represents a return to natural way of identification: testing someone by what (s)he is, instead relying on something owns or knows seems likely be the forward. Biometric systems that include multiple sources information are known as multimodal. Such generally regarded an alternative fight variety problems all unimodal stumble upon. One main challenges found in development biometric recognition is shortage publicly available databases acquired under real unconstrained working...
Facial expression recognition (FER) is currently one of the most active research topics due to its wide range applications in human-computer interaction field. An important part recent success automatic FER was achieved thanks emergence deep learning approaches. However, training networks for still a very challenging task, since available data sets are relatively small. Although transfer can partially alleviate issue, performance models below full potential as features may contain redundant...
The optical recognition of handwritten musical scores by computers remains far from ideal. Most OMR algorithms rely on an estimation the staff line thickness and vertical distance within same staff. Subsequent operation can use these values as references, dismissing need for some predetermined threshold values. In this work we improve previous conventional estimates two reference lengths. We start proposing a new method binarized music then extend approach gray-level scores. An experimental...
The detection of staff lines is the first step most Optical Music Recognition (OMR) systems. Its great significance derives from ease with which we can then proceed extraction musical symbols. All OMR tasks are usually achieved using binary images by setting thresholds that be local or global. These techniques however, may remove relevant information music sheet and introduce artifacts will degrade results in later stages process. It arises therefore a need to create method reduces loss due...
The preservation of many music works produced in the past entails their digitalization and consequent accessibility an easy-to-manage digital format. Carrying this task manually is very time consuming error prone. While optical recognition systems usually perform well on printed scores, processing handwritten musical scores by computers remain far from ideal. One fundamental stages to carry out staff line detection. In paper a new method for automatic detection lines based connected path...
As a key technology to help bridging the gap between deaf and hearing people, sign language recognition (SLR) has become one of most active research topics in human-computer interaction field. Although several SLR methodologies have been proposed, development real-world system is still very challenging task. One main challenges related large intersigner variability that exists manual signing process languages. To address this problem, we propose novel end-to-end deep neural network...
Many music works produced in the past still exist only as original manuscripts or photocopies. Preserving them entails their digitalization and consequent accessibility a digital format easy-to-manage. The manual process to carry out this task is very time consuming error prone. Optical recognition (OMR) form of structured document image analysis where symbols are isolated identified so that can be conveniently processed. While OMR systems perform well on printed scores, current methods for...
Although Optical Music Recognition (OMR) has been the focus of much research for decades, processing handwritten musical scores is not yet satisfactory. The efforts made to find robust symbol representations and learning methodologies have found a similar quality in dissimilarity concept. Simple Euclidean distances are often used measure between different examples. However, such do necessarily yield best performance. In this paper, we propose learn distance k-nearest neighbor (k-NN)...
As technology and artificial intelligence conquer a place under the spotlight in automotive world, driver drowsiness monitoring systems have sparked much interest as way to increase safety avoid sleepiness-related accidents. Such technologies, however, stumble upon observation that each presents distinct set of behavioral physiological manifestations drowsiness, thus rendering its objective assessment non-trivial process. The AUTOMOTIVE project studied application signal processing machine...
Many music works produced in the past still exist only as original manuscripts or photocopies. Preserving them entails their digitalization and consequent accessibility a digital format easy-to-manage. The manual process to carry out this task is very time consuming error prone. Optical recognition (OMR) form of structured document image analysis where symbols are isolated identified so that can be conveniently processed. While OMR systems perform well on printed scores, current methods for...
In this paper, a new method for music symbol classification named Combined Neural Network (CNN) is proposed. Tests are conducted on more than 9000 symbols from both real and scanned sheets, which show that the proposed technique offers superior capability. At same time, performance of network compared with single (NN) classifier using scores. The average accuracy increased ten percent, reaching 98.82%.
Optical Music Recognition (OMR) has received increasing attention in recent years. In this paper, we propose a classifier based on new method named Directed Acyclic Graph-Large margin Distribution Machine (DAG-LDM). The DAG-LDM is an improvement of the Large (LDM), which binary that optimizes distribution by maximizing mean and minimizing variance simultaneously. We modify LDM to solve multi-class music symbol classification problem. Tests are conducted more than 10000 images, obtained from...