Ferramentas e recursos livres para reconhecimento e síntese de voz em português brasileiro

Sistemas de reconhecimento e síntese de voz são constituídos por módulos que dependem da língua e, enquanto existem muitos recursos públicos para alguns idiomas (p.e. Inglês e Japonês), os recursos para Português Brasileiro (PB) ainda são escassos. Outro aspecto é que, para um grande número de tarefas, a taxa de erro dos sistemas de reconhecimento de voz atuais ainda é elevada, quando comparada à obtida por seres humanos. Assim, apesar do sucesso das cadeias escondidas de Markov (HMM), é necessária a pesquisa por novos métodos. Este trabalho tem como motivação esses dois fatos e se divide em duas partes. A primeira descreve o desenvolvimento de recursos e ferramentas livres para reconhecimento e síntese de voz em PB, consistindo de bases de dados de áudio e texto, um dicionário fonético, um conversor grafema-fone, um separador silábico e modelos acústico e de linguagem. Todos os recursos construídos encontram-se publicamente disponíveis e, junto com uma interface de programação proposta, têm sido usados para o desenvolvimento de várias novas aplicações em tempo-real, incluindo um módulo de reconhecimento de voz para a suíte de aplicativos para escritório OpenOffice.org. São apresentados testes de desempenho dos sistemas desenvolvidos. Os recursos aqui produzidos e disponibilizados facilitam a adoção da tecnologia de voz para PB por outros grupos de pesquisa, desenvolvedores e pela indústria. A segunda parte do trabalho apresenta um novo método para reavaliar (rescoring) o resultado do reconhecimento baseado em HMMs, o qual é organizado em uma estrutura de dados do tipo lattice. Mais especificamente, o sistema utiliza classificadores discriminativos que buscam diminuir a confusão entre pares de fones. Para cada um desses problemas binários, são usadas técnicas de seleção automática de parâmetros para escolher a representaçãao paramétrica mais adequada para o problema em questão.

Controle de qualidade para 3D-vídeo streaming em redes em malha sem fio

Redes em Malha sem Fio ( do inglês Wireless Mesh Networks - WMNs) são previstas serem uma das mais importantes tecnologias sem fio no que se refere ao fornecimento do acesso de última milha em redes multimídia futuras. Elas vão permitir que milhares de usuários fixos e móveis acessem, produzam e compartilhem conteúdo multimídia de forma onipresente. Neste contexto, vídeo 3D está previsto atrair mais e mais o mercado multimídia com a perspectiva de reforçar as aplicações (vídeos de vigilância, controle demissões críticas, entretenimento, etc). No entanto, o desafio de lidar com a largura de banda optante, escassez de recursos e taxas de erros variantes com o tempo destas redes, ilustra a necessidade da transmissão de vídeos 3D mais resistentes a erros. Dessa forma, alternativas como abordagens de Correção Antecipada de Erros (FEC) se tornam necessárias para fornecer a distribuição de aplicações de vídeo para usuários sem fio com garantia de melhor qualidade de serviço (QoS) e Qualidade de Experiência (QoE). Esta dissertação apresenta um mecanismo baseado em FEC com Proteção Desigual de Erros (UEP) para melhorar a transmissão de vídeo 3D em WMNs, aumentando a satisfação do usuário e permitindo uma melhoria do uso dos recursos sem fio. Os benefícios e impactos do mecanismo proposto serão demonstrados usando simulação e a avaliação será realizada através de métricas de QoE objetivas e subjetivas.

Identificação molecular dos peixes da bacia do alto rio Paraná

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Ambiente web de suporte à transcrição fonética automática de lemas em verbetes de dicionários do português do Brasil

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

A complex networks approach for data clustering

This work proposes a method for data clustering based on complex networks theory. A data set is represented as a network by considering different metrics to establish the connection between each pair of objects. The clusters are obtained by taking into account five community detection algorithms. The network-based clustering approach is applied in two real-world databases and two sets of artificially generated data. The obtained results suggest that the exponential of the Minkowski distance is the most suitable metric to quantify the similarities between pairs of objects. In addition, the community identification method based on the greedy optimization provides the best cluster solution. We compare the network-based clustering approach with some traditional clustering algorithms and verify that it provides the lowest classification error rate. (C) 2012 Elsevier B.V. All rights reserved.

Gender classification in large databases

[EN]In this paper, we address the challenge of gender classi - cation using large databases of images with two goals. The rst objective is to evaluate whether the error rate decreases compared to smaller databases. The second goal is to determine if the classi er that provides the best classi cation rate for one database, improves the classi cation results for other databases, that is, the cross-database performance.

Adaptive wireless multimedia communication systems

In recent years, due to the rapid convergence of multimedia services, Internet and wireless communications, there has been a growing trend of heterogeneity (in terms of channel bandwidths, mobility levels of terminals, end-user quality-of-service (QoS) requirements) for emerging integrated wired/wireless networks. Moreover, in nowadays systems, a multitude of users coexists within the same network, each of them with his own QoS requirement and bandwidth availability. In this framework, embedded source coding allowing partial decoding at various resolution is an appealing technique for multimedia transmissions. This dissertation includes my PhD research, mainly devoted to the study of embedded multimedia bitstreams in heterogenous networks, developed at the University of Bologna, advised by Prof. O. Andrisano and Prof. A. Conti, and at the University of California, San Diego (UCSD), where I spent eighteen months as a visiting scholar, advised by Prof. L. B. Milstein and Prof. P. C. Cosman. In order to improve the multimedia transmission quality over wireless channels, joint source and channel coding optimization is investigated in a 2D time-frequency resource block for an OFDM system. We show that knowing the order of diversity in time and/or frequency domain can assist image (video) coding in selecting optimal channel code rates (source and channel code rates). Then, adaptive modulation techniques, aimed at maximizing the spectral efficiency, are investigated as another possible solution for improving multimedia transmissions. For both slow and fast adaptive modulations, the effects of imperfect channel estimation errors are evaluated, showing that the fast technique, optimal in ideal systems, might be outperformed by the slow adaptive modulation, when a real test case is considered. Finally, the effects of co-channel interference and approximated bit error probability (BEP) are evaluated in adaptive modulation techniques, providing new decision regions concepts, and showing how the widely used BEP approximations lead to a substantial loss in the overall performance.

Context-aware wireless networks

This thesis investigates context-aware wireless networks, capable to adapt their behavior to the context and the application, thanks to the ability of combining communication, sensing and localization. Problems of signals demodulation, parameters estimation and localization are addressed exploiting analytical methods, simulations and experimentation, for the derivation of the fundamental limits, the performance characterization of the proposed schemes and the experimental validation. Ultrawide-bandwidth (UWB) signals are in certain cases considered and non-coherent receivers, allowing the exploitation of the multipath channel diversity without adopting complex architectures, investigated. Closed-form expressions for the achievable bit error probability of novel proposed architectures are derived. The problem of time delay estimation (TDE), enabling network localization thanks to ranging measurement, is addressed from a theoretical point of view. New fundamental bounds on TDE are derived in the case the received signal is partially known or unknown at receiver side, as often occurs due to propagation or due to the adoption of low-complexity estimators. Practical estimators, such as energy-based estimators, are revised and their performance compared with the new bounds. The localization issue is addressed with experimentation for the characterization of cooperative networks. Practical algorithms able to improve the accuracy in non-line-of-sight (NLOS) channel conditions are evaluated on measured data. With the purpose of enhancing the localization coverage in NLOS conditions, non-regenerative relaying techniques for localization are introduced and ad hoc position estimators are devised. An example of context-aware network is given with the study of the UWB-RFID system for detecting and locating semi-passive tags. In particular a deep investigation involving low-complexity receivers capable to deal with problems of multi-tag interference, synchronization mismatches and clock drift is presented. Finally, theoretical bounds on the localization accuracy of this and others passive localization networks (e.g., radar) are derived, also accounting for different configurations such as in monostatic and multistatic networks.

Einfluss der Einzeldosisblisterverpackung von festen Oralia auf die Arzneimitteltherapiesicherheit, Information und Zufriedenheit stationär behandelter Patienten in einer prospektiven, multizentrischen Studie

Bei der vorliegenden Studie wurde die Machbarkeit und Qualität der Arzneimittelverteilung von oralen Arzneimitteln in Einzeldosisblisterverpackungen je abgeteilte Arzneiform (EVA) untersucht.rnDie Studie wurde als offene, vergleichende, prospektive und multizentrische Patientenstudie durchgeführt. Als Studienmedikation standen Diovan®, CoDiovan® und Amlodipin in der EVA-Verpackung zur Verfügung. Die Verteilfehlerrate in der EVA- und Kontroll-Gruppe stellte den primären Zielparameter dar. Das Patientenwissen, die Patientenzufriedenheit und die Praktikabilität des EVA-Systems, sowie die Zufriedenheit der Pflegekräfte wurden mithilfe von Fragebogen evaluiert. Insgesamt wurden 2070 gültige Tablettenvergaben bei 332 Patienten in sechs verschiedenen Krankenhäusern geprüft. Es wurde in der EVA-Gruppe ein Verteilungsfehler von 1,8% und in der Kontroll-Gruppe von 0,7% ermittelt. Bei den Patienten-Fragebogen konnten insgesamt 292 Fragebogen ausgewertet werden. Die Ergebnisse zeigten einen ungenügenden Informationsstand der Patienten über ihre aktuellen, oralen Arzneimittel. In den 80 ausgefüllten Pflegekräfte-Fragebogen gaben über 80% an, dass Fehler beim Richten durch das EVA-System besser erkannt werden können. rnZusammenfassend kann gesagt werden, dass die erhöhte Fehlerrate in der EVA-Gruppe im Vergleich zur Kontroll-Gruppe durch mehrere Störfaktoren bedingt wurde. Grundsätzlich konnte eine sehr positive Resonanz auf das EVA-System bei den Patienten und den Pflegekräften beobachtet werden. rn

'Alternate-goal bias' in antisaccades and the influence of expectation

Saccadic performance depends on the requirements of the current trial, but also may be influenced by other trials in the same experiment. This effect of trial context has been investigated most for saccadic error rate and reaction time but seldom for the positional accuracy of saccadic landing points. We investigated whether the direction of saccades towards one goal is affected by the location of a second goal used in other trials in the same experimental block. In our first experiment, landing points ('endpoints') of antisaccades but not prosaccades were shifted towards the location of the alternate goal. This spatial bias decreased with increasing angular separation between the current and alternative goals. In a second experiment, we explored whether expectancy about the goal location was responsible for the biasing of the saccadic endpoint. For this, we used a condition where the saccadic goal randomly changed from one trial to the next between locations on, above or below the horizontal meridian. We modulated the prior probability of the alternate-goal location by showing cues prior to stimulus onset. The results showed that expectation about the possible positions of the saccadic goal is sufficient to bias saccadic endpoints and can account for at least part of this phenomenon of 'alternate-goal bias'.

Why do humans make antisaccade errors?

Antisaccade errors are attributed to failure to inhibit the habitual prosaccade. We investigated whether the amount of information about the required response the patient has before the trial begins also contributes to error rate. Participants performed antisaccades in five conditions. The traditional design had two goals on the left and right horizontal meridians. In the second condition, stimulus-goal confusability between trials was eliminated by displacing one goal upward. In the third, hemifield uncertainty was eliminated by placing both goals in the same hemifield. In the fourth, goal uncertainty was eliminated by having only one goal, but interspersed with no-go trials. The fifth condition eliminated all uncertainty by having the same goal on every trial. Antisaccade error rate increased by 2% with each additional source of uncertainty, with the main effect being hemifield information, and a trend for stimulus-goal confusability. A control experiment for the effects of increasing angular separation between targets without changing these types of prior response information showed no effects on latency or error rate. We conclude that other factors besides prosaccade inhibition contribute to antisaccade error rates in traditional designs, possibly by modulating the strength of goal activation.

A NOVEL AND SIMPLE RULE OF THUMB FOR MULTIPLICITY CONTROL IN EQUIVALENCE TESTING USING TWO ONE-SIDED TESTS

Equivalence testing is growing in use in scientific research outside of its traditional role in the drug approval process. Largely due to its ease of use and recommendation from the United States Food and Drug Administration guidance, the most common statistical method for testing (bio)equivalence is the two one-sided tests procedure (TOST). Like classical point-null hypothesis testing, TOST is subject to multiplicity concerns as more comparisons are made. In this manuscript, a condition that bounds the family-wise error rate (FWER) using TOST is given. This condition then leads to a simple solution for controlling the FWER. Specifically, we demonstrate that if all pairwise comparisons of k independent groups are being evaluated for equivalence, then simply scaling the nominal Type I error rate down by (k - 1) is sufficient to maintain the family-wise error rate at the desired value or less. The resulting rule is much less conservative than the equally simple Bonferroni correction. An example of equivalence testing in a non drug-development setting is given.

Beitrag zur Dimensionierung von automatischen Sortiersystemen (Sorter)

Automatischen Sortiersysteme (Sorter) besitzen in der Intralogistik eine große Bedeutung. Sorter erreichen eine ausdauernd hohe Sortierleistung bei gleichzeitig geringer Fehlsortierrate und bilden deshalb oft den zentralen Baustein in Materialflusssystemen mit hoher Umschlagsrate. Distributionszentren mit Lager und Kommissionierfunktion sind typische Vertreter solcher Materialflusssysteme. Ein Sorter besteht aus den Subsystemen Einschleusung, Verteilförderer und Endstellen. Die folgenden Betrachtungen fokussieren auf ein Sortermodell mit einem Verteilförderer in Ringstruktur und einer Einzelplatzbelegung. Auf jedem Platz kann genau ein Gut transportiert werden. Der Verteilförderer besitzt somit eine feste Transportkapazität. Derartige Förderer werden in der Regel als Kippschalen- oder Quergurt-Sorter ausgeführt. Die theoretische Sortierleistung für diesen Sortertyp kann aus Fahrgeschwindigkeit und Transportplatzabstand bestimmt werden. Diese Systemleistung wird im praktischen Betrieb kaum erreicht. Verschiedene Faktoren im Einschleusbereich und im Ausschleusbereich führen zu einer Leistungsminderung. Betrachtungen zur Bestimmung der mittleren Warteschlangenlänge im Einschleusbereich sowie zur Ermittlung des Rundläuferanteils auf dem Verteilförderer werden im folgenden Beitrag vorgestellt. Diesem Beitrag liegt ein Forschungsvorhaben zugrunde, das aus Mitteln des Bundesministeriums für Wirtschaft und Technologie (BMWi) über die Arbeitsgemeinschaft industrieller Forschungsvereinigungen "Otto von Guericke" (AiF) gefördert und im Auftrage der Bundesvereinigung Logistik e.V. (BVL) ausgeführt wurde.

Kommissionierung par excellence

Der Kommissionierprozess stellt im Rahmen der innerbetrieblichen Logistik - gerade auch im Hinblick auf Just-In-Time-Lieferungen und Fragen der Produkthaftung - einen zentralen Baustein des Material- und Informationsflusses in Unternehmen dar. Dabei ist die Wahl des Kommissioniersystems ausschlaggebend für die Optimierung der personal- und zeitaufwendigen Kommissioniervorgänge und dient damit zur Leistungssteigerung unter gleichzeitiger Reduzierung der Fehlerquote.