On the determination of epsilon during discriminative GMM training

Discriminative training of Gaussian Mixture Models (GMMs) for speech or speaker recognition purposes is usually based on the gradient descent method, in which the iteration step-size, ε, uses to be defined experimentally. In this letter, we derive an equation to adaptively determine ε, by showing that the second-order Newton-Raphson iterative method to find roots of equations is equivalent to the gradient descent algorithm. © 2010 IEEE.

Distributed generators as providers of reactive power support - A market approach

Traditionally, ancillary services are supplied by large conventional generators. However, with the huge penetration of distributed generators (DGs) as a result of the growing interest in satisfying energy requirements, and considering the benefits that they can bring along to the electrical system and to the environment, it appears reasonable to assume that ancillary services could also be provided by DGs in an economical and efficient way. In this paper, a settlement procedure for a reactive power market for DGs in distribution systems is proposed. Attention is directed to wind turbines connected to the network through synchronous generators with permanent magnets and doubly-fed induction generators. The generation uncertainty of this kind of DG is reduced by running a multi-objective optimization algorithm in multiple probabilistic scenarios through the Monte Carlo method and by representing the active power generated by the DGs through Markov models. The objectives to be minimized are the payments of the distribution system operator to the DGs for reactive power, the curtailment of transactions committed in an active power market previously settled, the losses in the lines of the network, and a voltage profile index. The proposed methodology was tested using a modified IEEE 37-bus distribution test system. © 1969-2012 IEEE.

Análise de técnicas de reconhecimento de padrões para a identificação biométrica de usuários em aplicações WEB Utilizando faces a partir de vídeos

Pós-graduação em Ciência da Computação - IBILCE

Conversão grafema-fone para um sistema de reconhecimento de voz com suporte a grandes vocabulários para o português brasileiro

O processamento de voz tornou-se uma tecnologia cada vez mais baseada na modelagem automática de vasta quantidade de dados. Desta forma, o sucesso das pesquisas nesta área está diretamente ligado a existência de corpora de domínio público e outros recursos específicos, tal como um dicionário fonético. No Brasil, ao contrário do que acontece para a língua inglesa, por exemplo, não existe atualmente em domínio público um sistema de Reconhecimento Automático de Voz (RAV) para o Português Brasileiro com suporte a grandes vocabulários. Frente a este cenário, o trabalho tem como principal objetivo discutir esforços dentro da iniciativa FalaBrasil [1], criada pelo Laboratório de Processamento de Sinais (LaPS) da UFPA, apresentando pesquisas e softwares na área de RAV para o Português do Brasil. Mais especificamente, o presente trabalho discute a implementação de um sistema de reconhecimento de voz com suporte a grandes vocabulários para o Português do Brasil, utilizando a ferramenta HTK baseada em modelo oculto de Markov (HMM) e a criação de um módulo de conversão grafema-fone, utilizando técnicas de aprendizado de máquina.

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.

Reconhecimento de voz para aplicações em automação implementado em FPGA

In many movies of scientific fiction, machines were capable of speaking with humans. However mankind is still far away of getting those types of machines, like the famous character C3PO of Star Wars. During the last six decades the automatic speech recognition systems have been the target of many studies. Throughout these years many technics were developed to be used in applications of both software and hardware. There are many types of automatic speech recognition system, among which the one used in this work were the isolated word and independent of the speaker system, using Hidden Markov Models as the recognition system. The goals of this work is to project and synthesize the first two steps of the speech recognition system, the steps are: the speech signal acquisition and the pre-processing of the signal. Both steps were developed in a reprogrammable component named FPGA, using the VHDL hardware description language, owing to the high performance of this component and the flexibility of the language. In this work it is presented all the theory of digital signal processing, as Fast Fourier Transforms and digital filters and also all the theory of speech recognition using Hidden Markov Models and LPC processor. It is also presented all the results obtained for each one of the blocks synthesized e verified in hardware

Using Threads to Overcome Synchronization Delays in Parallel Multiple Progressive Alignment Algorithms

Sao Paulo State Research Foundation-FAPESP

Learning HMMs for nucleotide sequences from amino acid alignments

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

Statistical methods for the analysis of DNA sequences: application to dinucleotide distribution in the human genome

Questa tesi si inserisce nell'ambito delle analisi statistiche e dei metodi stocastici applicati all'analisi delle sequenze di DNA. Nello specifico il nostro lavoro è incentrato sullo studio del dinucleotide CG (CpG) all'interno del genoma umano, che si trova raggruppato in zone specifiche denominate CpG islands. Queste sono legate alla metilazione del DNA, un processo che riveste un ruolo fondamentale nella regolazione genica. La prima parte dello studio è dedicata a una caratterizzazione globale del contenuto e della distribuzione dei 16 diversi dinucleotidi all'interno del genoma umano: in particolare viene studiata la distribuzione delle distanze tra occorrenze successive dello stesso dinucleotide lungo la sequenza. I risultati vengono confrontati con diversi modelli nulli: sequenze random generate con catene di Markov di ordine zero (basate sulle frequenze relative dei nucleotidi) e uno (basate sulle probabilità di transizione tra diversi nucleotidi) e la distribuzione geometrica per le distanze. Da questa analisi le proprietà caratteristiche del dinucleotide CpG emergono chiaramente, sia dal confronto con gli altri dinucleotidi che con i modelli random. A seguito di questa prima parte abbiamo scelto di concentrare le successive analisi in zone di interesse biologico, studiando l’abbondanza e la distribuzione di CpG al loro interno (CpG islands, promotori e Lamina Associated Domains). Nei primi due casi si osserva un forte arricchimento nel contenuto di CpG, e la distribuzione delle distanze è spostata verso valori inferiori, indicando che questo dinucleotide è clusterizzato. All’interno delle LADs si trovano mediamente meno CpG e questi presentano distanze maggiori. Infine abbiamo adottato una rappresentazione a random walk del DNA, costruita in base al posizionamento dei dinucleotidi: il walk ottenuto presenta caratteristiche drasticamente diverse all’interno e all’esterno di zone annotate come CpG island. Riteniamo pertanto che metodi basati su questo approccio potrebbero essere sfruttati per migliorare l’individuazione di queste aree di interesse nel genoma umano e di altri organismi.

Processi markoviani nascosti

We introduce the notation of Markov chains and their properties, and give the definition of ergodic, irreducible and aperiodic chains with correspective examples. Then, the definition of hidden Markov models is given and their characteristics are examined. We formulate three basic problems regarding the hidden Markov models and discuss the solution of two of them - the Viterbi algorithm and the forward-backward algorithm.

Técnicas de análisis, caracterización y detección de señales de voz en entornos acústicos adversos

Este trabajo de Tesis ha abordado el objetivo de dar robustez y mejorar la Detección de Actividad de Voz en entornos acústicos adversos con el fin de favorecer el comportamiento de muchas aplicaciones vocales, por ejemplo aplicaciones de telefonía basadas en reconocimiento automático de voz, aplicaciones en sistemas de transcripción automática, aplicaciones en sistemas multicanal, etc. En especial, aunque se han tenido en cuenta todos los tipos de ruido, se muestra especial interés en el estudio de las voces de fondo, principal fuente de error de la mayoría de los Detectores de Actividad en la actualidad. Las tareas llevadas a cabo poseen como punto de partida un Detector de Actividad basado en Modelos Ocultos de Markov, cuyo vector de características contiene dos componentes: la energía normalizada y la variación de la energía. Las aportaciones fundamentales de esta Tesis son las siguientes: 1) ampliación del vector de características de partida dotándole así de información espectral, 2) ajuste de los Modelos Ocultos de Markov al entorno y estudio de diferentes topologías y, finalmente, 3) estudio e inclusión de nuevas características, distintas de las del punto 1, para filtrar los pulsos de pronunciaciones que proceden de las voces de fondo. Los resultados de detección, teniendo en cuenta los tres puntos anteriores, muestran con creces los avances realizados y son significativamente mejores que los resultados obtenidos, bajo las mismas condiciones, con otros detectores de actividad de referencia. This work has been focused on improving the robustness at Voice Activity Detection in adverse acoustic environments in order to enhance the behavior of many vocal applications, for example telephony applications based on automatic speech recognition, automatic transcription applications, multichannel systems applications, and so on. In particular, though all types of noise have taken into account, this research has special interest in the study of pronunciations coming from far-field speakers, the main error source of most activity detectors today. The tasks carried out have, as starting point, a Hidden Markov Models Voice Activity Detector which a feature vector containing two components: normalized energy and delta energy. The key points of this Thesis are the following: 1) feature vector extension providing spectral information, 2) Hidden Markov Models adjustment to environment and study of different Hidden Markov Model topologies and, finally, 3) study and inclusion of new features, different from point 1, to reject the pronunciations coming from far-field speakers. Detection results, taking into account the above three points, show the advantages of using this method and are significantly better than the results obtained under the same conditions by other well-known voice activity detectors.

Analysis of pattern recognition techniques for in-air signature biometrics

As a result of advances in mobile technology, new services which benefit from the ubiquity of these devices are appearing. Some of these services require the identification of the subject since they may access private user information. In this paper, we propose to identify each user by drawing his/her handwritten signature in the air (in-airsignature). In order to assess the feasibility of an in-airsignature as a biometric feature, we have analysed the performance of several well-known patternrecognitiontechniques—Hidden Markov Models, Bayes classifiers and dynamic time warping—to cope with this problem. Each technique has been tested in the identification of the signatures of 96 individuals. Furthermore, the robustness of each method against spoofing attacks has also been analysed using six impostors who attempted to emulate every signature. The best results in both experiments have been reached by using a technique based on dynamic time warping which carries out the recognition by calculating distances to an average template extracted from several training instances. Finally, a permanence analysis has been carried out in order to assess the stability of in-airsignature over time.

HF spectrum activity prediction model based on HMM for cognitive radio applications

Although most of the research on Cognitive Radio is focused on communication bands above the HF upper limit (30 MHz), Cognitive Radio principles can also be applied to HF communications to make use of the extremely scarce spectrum more efficiently. In this work we consider legacy users as primary users since these users transmit without resorting to any smart procedure, and our stations using the HFDVL (HF Data+Voice Link) architecture as secondary users. Our goal is to enhance an efficient use of the HF band by detecting the presence of uncoordinated primary users and avoiding collisions with them while transmitting in different HF channels using our broad-band HF transceiver. A model of the primary user activity dynamics in the HF band is developed in this work to make short-term predictions of the sojourn time of a primary user in the band and avoid collisions. It is based on Hidden Markov Models (HMM) which are a powerful tool for modelling stochastic random processes and are trained with real measurements of the 14 MHz band. By using the proposed HMM based model, the prediction model achieves an average 10.3% prediction error rate with one minute-long channel knowledge but it can be reduced when this knowledge is extended: with the previous 8 min knowledge, an average 5.8% prediction error rate is achieved. These results suggest that the resulting activity model for the HF band could actually be used to predict primary users activity and included in a future HF cognitive radio based station.

Efficiency improvement of HF communications using cognitive radio principles

Cognitive Radio principles can be applied to HF communications to make a more efficient use of the extremely scarce spectrum. In this contribution we focus on analyzing the usage of the available channels done by the legacy users, which are regarded as primary users since they are allowed to transmit without resorting any smart procedure, and consider the possibilities for our stations -over the HFDVL (HF Data+Voice Link) architecture- to participate as secondary users. Our goal is to enhance an efficient use of the HF band by detecting the presence of uncoordinated primary users and avoiding collisions with them while transmitting in different HF channels using our broad-band HF transceiver. A model of the primary user activity dynamics in the HF band is developed in this work. It is based on Hidden Markov Models (HMM) which are a powerful tool for modelling stochastic random processes, and is trained with real measurements from the 14 MHz band.

Propuesta, implementación y evaluación de la biometría de firma en el aire como sistema de verificación en teléfonos móviles

En esta tesis doctoral se propone una técnica biométrica de verificación en teléfonos móviles consistente en realizar una firma en el aire con la mano que sujeta el teléfono móvil. Los acelerómetros integrados en el dispositivo muestrean las aceleraciones del movimiento de la firma en el aire, generando tres señales temporales que pueden utilizarse para la verificación del usuario. Se proponen varios enfoques para la implementación del sistema de verificación, a partir de los enfoques más utilizados en biometría de firma manuscrita: correspondencia de patrones, con variantes de los algoritmos de Needleman-Wusch (NW) y Dynamic Time Warping (DTW), modelos ocultos de Markov (HMM) y clasificador estadístico basado en Máquinas de Vector Soporte (SVM). Al no existir bases de datos públicas de firmas en el aire y con el fin de evaluar los métodos propuestos en esta tesis doctoral, se han capturado dos con distintas características; una con falsificaciones reales a partir del estudio de las grabaciones de usuarios auténticos y otra con muestras de usuarios obtenidas en diferentes sesiones a lo largo del tiempo. Utilizando estas bases de datos se han evaluado una gran cantidad de algoritmos para implementar un sistema de verificación basado en firma en el aire. Esta evaluación se ha realizado de acuerdo con el estándar ISO/IEC 19795, añadiendo el caso de verificación en mundo abierto no incluido en la norma. Además, se han analizado las características que hacen que una firma sea suficientemente segura. Por otro lado, se ha estudiado la permanencia de las firmas en el aire a lo largo del tiempo, proponiendo distintos métodos de actualización, basados en una adaptación dinámica del patrón, para mejorar su rendimiento. Finalmente, se ha implementado un prototipo de la técnica de firma en el aire para teléfonos Android e iOS. Los resultados de esta tesis doctoral han tenido un gran impacto, generando varias publicaciones en revistas internacionales, congresos y libros. La firma en el aire ha sido nombrada también en varias revistas de divulgación, portales de noticias Web y televisión. Además, se han obtenido varios premios en competiciones de ideas innovadoras y se ha firmado un acuerdo de explotación de la tecnología con una empresa extranjera. ABSTRACT This thesis proposes a biometric verification technique on mobile phones consisting on making a signature in the air with the hand holding a mobile phone. The accelerometers integrated in the device capture the movement accelerations, generating three temporal signals that can be used for verification. This thesis suggests several approaches for implementing the verification system, based on the most widely used approaches in handwritten signature biometrics: template matching, with a lot of variations of the Needleman- Wusch (NW) and Dynamic Time Warping (DTW) algorithms, Hidden Markov Models (HMM) and Supported Vector Machines (SVM). As there are no public databases of in-air signatures and with the aim of assessing the proposed methods, there have been captured two databases; one. with real falsification attempts from the study of recordings captured when genuine users made their signatures in front of a camera, and other, with samples obtained in different sessions over a long period of time. These databases have been used to evaluate a lot of algorithms in order to implement a verification system based on in-air signatures. This evaluation has been conducted according to the standard ISO/IEC 19795, adding the open-set verification scenario not included in the norm. In addition, the characteristics of a secure signature are also investigated, as well as the permanence of in-air signatures over time, proposing several updating strategies to improve its performance. Finally, a prototype of in-air signature has been developed for iOS and Android phones. The results of this thesis have achieved a high impact, publishing several articles in SCI journals, conferences and books. The in-air signature deployed in this thesis has been also referred in numerous media. Additionally, this technique has won several awards in the entrepreneurship field and also an exploitation agreement has been signed with a foreign company.