Carlos Henrique Quartucci Forster

Analysis of aircraft trajectories using Fourier descriptors and kernel density estimation

Handling the large amount of information from aircraft trajectories that are produced daily from air traffic control radar systems requires models for representing trajectories in a compact, easy to calculate, representative and distinctive form. These models should permit to perform clustering and classification operations efficiently and effectively. The Fourier descriptors have these characteristics and this article presents the results obtained on actual aircraft trajectories including approach and takeoff operations over a terminal area. Clustering and classification techniques in the feature space of Fourier descriptors were able to correctly separate the various types of operations. Additionally, based on the results of the clusters obtained, a method is presented for the classification of trajectories in progress based on kernel density estimation. An interesting result from the point of view of air traffic control for the detection of anomalous traffic is demonstrated.

TeleEye: An Awareness Widget for Providing the Focus of Attention in Collaborative Editing Systems

Awareness is the knowledge about present and past group’s activities and it is a relevant issue for cooperative work. There are many devices that supply awareness information in synchronous collaborative editing systems. However, the current awareness devices have restrictions to both accomplish effective awareness and show the focus of attention identifying the exact place of the participants’ attention. This paper presents an awareness widget for synchronous collaborative editing systems called TeleEye that provides information about the localization of the participants’ attention during a collaborative session by means of eye tracking.

Experimental comparison of DWT and DFT for trajectory representation

In this work Discrete Fourier Transform (DFT) and Discrete Wavelet Transform (DWT) were experimentally evaluated for their performances as tools for dimensionality reduction in a real data set of air traffic trajectories. Results showed that both DFT and DWT were able to provide very expressive reduction for trajectory representation with minimal loss of information. Overall, DWT performed slightly better requiring fewer coefficients than DFT to achieve the same signal energy or to provide the same quality of reconstruction of the trajectories.

Supervised growing neural gas

We present a new approach to supervised vector quantization inspired on growing neural gas network. An advantage of the new method is that it reduces the need for prior knowledge about the problem under study because it is able to determine at runtime the size of the codebook. Another advantage is that the training is less dependent on the initial state of the codebook vectors in contrast to methods like Learning Vector Quantization. Finally, it is shown that for some real datasets the classification performance is superior to other methods of supervised vector quantization.

Color image segmentation using gaussian mixtures and particle swarm optimization

The model of Gaussian Mixture is particularly useful to perform unsupervised learning. Currently, the principal technique to estimate the mixture parameters is the Expectation Maximization method which has a great chance of obtaining sub-optimal results. In this work we opted, instead, for the Particle Swarm Optimization as an alternative way to estimate parameter of Gaussian Mixture applied to multivariate data, which has greater chance of reaching the optimum. To evaluate the proposed approach, color images from fluorescence microscopy are segmented considering the 3D color space. Some particular features of this kind of color image are also considered to improve the performance of the search.

Development of Human Faces Retrieval in a Big Photo Database with SCRUM: A Case Study

In this paper we present a case study of the use of SCRUM in an applied research scenario, consisting of the detection and identification of human faces in a large database of photos. The study shows how the adoption of agile methods was important to deal with uncertainty and unstable requirements, allowing adaptable development with opportunity to learn from the development process. The implementation of experiments reveals, as our research advances, the applicability of techniques to support the solution for the requirements. All sort of problems with illumination, pose, low quality of image and other type of noise were present as obstacles to accuracy in different approaches for detection and recognition. Fast retrieval of similar faces was also necessary. In the end, the requirements were successfully tackled by developing an efficient and short implementation of face alignment, feature extraction and search.

Um Experimento sobre o Emprego de TiddlyWiki como Mídia para Aulas e Exercícios na Área de Exatas

Resumo Este artigo apresenta um experimento realizado num curso de Calculo Numerico onde foi utilizada a TiddlyWiki em diversos contextos: como slides em sala de aula, como roteiro de exercicios de laboratorio e como formato para os relatorios a serem elaborados pelos alunos. Para que fosse aplicavel no curso, foi necessario buscar, adaptar e ate mesmo recodificar algumas extensoes da TiddlyWiki. Esta nova tecnologia de publicacao mostrou-se capaz de resolver alguns problemas existentes nas tecnologias atuais, como a dificuldade de publicacao, a dificuldade de estruturacao das ideias e a necessidade de aplicativos especificos para acesso ao material. Por outro lado, a tecnologia, ainda nao madura, apresentou problemas como a dificuldade de insercao de elementos nao-textuais e a complexidade do entendimento da estrutura de hiper-texto. Palavras-Chave: autoria hipermidia, tecnologias de publicacao, autoria colaborativa, Wiki, tecnologias de apoio em sala de aula

Desenvolvimento e avaliação de um sistema multimodal e multiusuário de navegação web

This paper presents the development and empirical evaluation of a voice-based web browser called BrowserVox, capable of dealing with different Brazilian-Portuguese websites. The system uses Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) components, providing opportunity for mouse or voice interaction (multimodal), with audiovisual feedback. With multiuser capability, it works immediately, without requires previous training of the speech recognition component. We get these features through the use of an auxiliary grammar which contains recognition rules dynamically created for each URL accessed. Even with the limited set of recognizable commands, we can create variations of these commands by using additional auxiliary words (eg please), in static and recursive rules included in grammar. This technique allows the commands to become more humanized. It also allows the system to ignore certain words that are uttered spontaneously by users during the dictation of the command, thus improving the efficiency of the speech recognition system. Finally, we applied usability testing, adopting a set of criteria proposed by Nielsen to evaluate the performance of the navigation system.

Utilização do Rastreamento Ocular para Visualização do Local de Atenção em Sistemas de Edição Colaborativos

Awareness is the knowledge about the present and past grouppsilas activities and it is a relevant question for cooperative work. There are many devices that supplies awareness information in synchronous collaborative editing systems. However, the current awareness devices have restrictions to accomplish effective awareness and to capture the focus of attention, e.g. identify the exact place of the participantspsila attentions. This paper presents an awareness device for synchronous collaborative editing systems called TeleEye that provide information about the place of the participantpsilas attentions during a collaborative session by means of eye tracking.

Evaluation of a multi-user system of voice interaction using grammars

This paper shows an experimental study about the design of grammars for a voice interface system. The influence of the grammar design on the behavior of the voice recognition system regarding accuracy and computational cost is assessed through tests. With the redesign of a grammar we show that those characteristics can be expressively improved.