No description available.

Speech is by far the most natural form of communication between human beings. It is intuitive, expressive and contains information at several cognitive levels. We as humans, are perceptive to several of these cognitive levels of information, as we can gather the information pertaining to the identity of the speaker, the speaker's gender, emotion, location, the language, and so on, in addition to the content of what is being spoken. This makes speech based human machine interaction (HMI), both desirable and challenging for the same set of reasons. For HMI to be natural for humans, it is imperative that a machine understands information present in speech, at least at the level of speaker identity, language, location in space, and the summary of what is being spoken. Although one can draw parallels between the human-human interaction and HMI, the two differ in their purpose. We, as humans, interact with a machine, mostly in the context of getting a task done more efficiently, than is possible without the machine. Thus, typically in HMI, controlling the machine in a specific manner is the primary goal. In this context, it can be argued that, HMI, with a limited vocabulary containing specific commands, would suffice for a more efficient use of the machine. In this thesis, we address the problem of ``Who spoke what and where", in the context of a machine understanding the information pertaining to identities of the speakers, their locations in space and the keywords they spoke, thus considering three levels of information - speaker identity (who), location (where) and keywords (what). This can be addressed with the help of multiple sensors like microphones, video camera, proximity sensors, motion detectors, etc., and combining all these modalities. However, we explore the use of only microphones to address this issue. In practical scenarios, often there are times, wherein, multiple people are talking at the same time. Thus, the goal of this thesis is to detect all the speakers, their keywords, and their locations in mixture signals containing speech from simultaneous speakers. Addressing this problem of ``Who spoke what and where" using only microphone signals, forms a part of acoustic scene analysis (ASA) of speech based acoustic events. We divide the problem of ``who spoke what and where" into two sub-problems: ``Who spoke what?" and ``Who spoke where". Each of these problems is cast in a generic latent variable (LV) framework to capture information in speech at different levels. We associate a LV to represent each of these levels and model the relationship between the levels using conditional dependency. The sub-problem of ``who spoke what" is addressed using single channel microphone signal, by modeling the mixture signal in terms of LV mass functions of speaker identity, the conditional mass function of the keyword spoken given the speaker identity, and a speaker-specific-keyword model. The LV mass functions are estimated in a Maximum likelihood (ML) framework using the Expectation Maximization (EM) algorithm using Student's-t Mixture Model (tMM) as speaker-specific-keyword models. Motivated by HMI in a home environment, we have created our own database. In mixture signals, containing two speakers uttering the keywords simultaneously, the proposed framework achieves an accuracy of 82 % for detecting both the speakers and their respective keywords. The other sub-problem of ``who spoke where?" is addressed in two stages. In the first stage, the enclosure is discretized into sectors. The speakers and the sectors in which they are located are detected in an approach similar to the one employed for ``who spoke what" using signals collected from a Uniform Circular Array (UCA). However, in place of speaker-specific-keyword models, we use tMM based speaker models trained on clean speech, along with a simple Delay and Sum Beamformer (DSB). In the second stage, the speakers are localized within the active sectors using a novel region constrained localization technique based on time difference of arrival (TDOA). Since the problem being addressed is a multi-label classification task, we use the average Hamming score (accuracy) as the performance metric. Although the proposed approach yields an accuracy of 100 % in an anechoic setting for detecting both the speakers and their corresponding sectors in two-speaker mixture signals, the performance degrades to an accuracy of 67 % in a reverberant setting, with a $60$ dB reverberation time (RT60) of 300 ms. To improve the performance under reverberation, prior knowledge of the location of multiple sources is derived using a novel technique derived from geometrical insights into TDOA estimation. With this prior knowledge, the accuracy of the proposed approach improves to 91 %. It is worthwhile to note that, the accuracies are computed for mixture signals containing more than 90 % overlap of competing speakers. The proposed LV framework offers a convenient methodology to represent information at broad levels. In this thesis, we have shown its use with three different levels. This can be extended to several such levels to be applicable for a generic analysis of the acoustic scene consisting of broad levels of events. It will turn out that not all levels are dependent on each other and hence the LV dependencies can be minimized by independence assumption, which will lead to solving several smaller sub-problems, as we have shown above. The LV framework is also attractive to incorporate prior knowledge about the acoustic setting, which is combined with the evidence from the data to derive the information about the presence of an acoustic event. The performance of the framework, is dependent on the choice of stochastic models, which model the likelihood function of the data given the presence of acoustic events. However, it provides an access to compare and contrast the use of different stochastic models for representing the likelihood function.

The migratory birds stop at different stopover sites during migration. The presence of resources in these stopover sites is essential to regain the energy of these birds. This thesis aims to compare the resource abundance and intake at the two stopover sites: Reda and Wisla river estuaries. How a bird's mass changes during its stay at an estuary is considered as a proxy for the resource abundance of a site. The comparison is made on different subsets, including those which has incomplete data, i.e. next day is not exactly one day after the previous capture. Multiple linear regression, Generalized additive model and Linear mixed effect model are used for analysis. Expectation maximization and an iterative predictive process are implemented to deal with incomplete data. We found that Reda has higher resource abundance and intake as compared to that of Wisla river estuary.

This master's thesis is dedicated to theme of knowledge discovery in Multimedia Databases, especially basic methods of classification and prediction used for data mining. The other part described about extraction of low level features from video data and images and summarizes information about content-based search in multimedia content and indexing this type of data. Final part is dedicated to implementation Gaussian mixtures model for classification and compare the final result with other method SVM.

This thesis present a Stochastic Volatility in Mean (SVM) model which is estimated using sequential Monte Carlo methods. The SVM model was first introduced by Koopman and provides an opportunity to study the intertemporal relationship between stock returns and their volatility through inclusion of volatility itself as an explanatory variable in the mean-equation. Using sequential Monte Carlo methods allows us to consider a non-linear estimation procedure at cost of introducing extra computational complexity. The recently developed PaRIS-algorithm, introduced by Olsson and Westerborn, drastically decrease the computational complexity of smoothing relative to previous algorithms and allows for efficient estimation of parameters. The main purpose of this thesis is to investigate the volatility feedback effect, i.e. the relation between expected return and unexpected volatility in an empirical study. The results shows that unanticipated shocks to the return process do not explain expected returns. / Detta examensarbete presenterar en stokastisk volatilitets medelvärdes (SVM) modell som estimeras genom sekventiella Monte Carlo metoder. SVM-modellen introducerades av Koopman och ger en möjlighet att studera den samtida relationen mellan aktiers avkastning och deras volatilitet genom att inkludera volatilitet som en förklarande variabel i medelvärdes-ekvationen. Sekventiella Monte Carlo metoder tillåter oss att använda icke-linjära estimerings procedurer till en kostnad av extra beräkningskomplexitet. Den nyligen utvecklad PaRIS-algoritmen, introducerad av Olsson och Westerborn, minskar drastiskt beräkningskomplexiteten jämfört med tidigare algoritmer och tillåter en effektiv uppskattning av parametrar. Huvudsyftet med detta arbete är att undersöka volatilitets-återkopplings-teorin d.v.s. relationen mellan förväntad avkastning och oväntad volatilitet i en empirisk studie. Resultatet visar på att oväntade chockar i avkastningsprocessen inte har förklarande förmåga över förväntad avkastning.

In radiation therapy treatment planning, recent works have used machine learning based on historically delivered plans to automate the process of producing clinically acceptable plans. Compared to traditional approaches such as repeated weighted-sum optimization or multicriteria optimization (MCO), automated planning methods have, in general, the benefits of low computational times and minimal user interaction, but on the other hand lack the flexibility associated with general-purpose frameworks such as MCO. Machine learning approaches can be especially sensitive to deviations in their dose prediction due to certain properties of the optimization functions usually used for dose mimicking and, moreover, suffer from the fact that there exists no general causality between prediction accuracy and optimized plan quality.In this thesis, we present a means of unifying ideas from machine learning planning methods with the well-established MCO framework. More precisely, given prior knowledge in the form of either a previously optimized plan or a set of historically delivered clinical plans, we are able to automatically generate Pareto optimal plans spanning a dose region corresponding to plans which are achievable as well as clinically acceptable. For the former case, this is achieved by introducing dose--volume constraints; for the latter case, this is achieved by fitting a weighted-data Gaussian mixture model on pre-defined dose statistics using the expectation--maximization algorithm, modifying it with exponential tilting and using specially developed optimization functions to take into account prediction uncertainties.Numerical results for conceptual demonstration are obtained for a prostate cancer case with treatment delivered by a volumetric-modulated arc therapy technique, where it is shown that the methods developed in the thesis are successful in automatically generating Pareto optimal plans of satisfactory quality and diversity, while excluding clinically irrelevant dose regions. For the case of using historical plans as prior knowledge, the computational times are significantly shorter than those typical of conventional MCO. / Inom strålterapiplanering har den senaste forskningen använt maskininlärning baserat på historiskt levererade planer för att automatisera den process i vilken kliniskt acceptabla planer produceras. Jämfört med traditionella angreppssätt, såsom upprepad optimering av en viktad målfunktion eller flermålsoptimering (MCO), har automatiska planeringsmetoder generellt sett fördelarna av lägre beräkningstider och minimal användarinteraktion, men saknar däremot flexibiliteten hos allmänna ramverk som exempelvis MCO. Maskininlärningsmetoder kan vara speciellt känsliga för avvikelser i dosprediktionssteget på grund av särskilda egenskaper hos de optimeringsfunktioner som vanligtvis används för att återskapa dosfördelningar, och lider dessutom av problemet att det inte finns något allmängiltigt orsakssamband mellan prediktionsnoggrannhet och kvalitet hos optimerad plan. I detta arbete presenterar vi ett sätt att förena idéer från maskininlärningsbaserade planeringsmetoder med det väletablerade MCO-ramverket. Mer precist kan vi, givet förkunskaper i form av antingen en tidigare optimerad plan eller en uppsättning av historiskt levererade kliniska planer, automatiskt generera Paretooptimala planer som täcker en dosregion motsvarande uppnåeliga såväl som kliniskt acceptabla planer. I det förra fallet görs detta genom att introducera dos--volym-bivillkor; i det senare fallet görs detta genom att anpassa en gaussisk blandningsmodell med viktade data med förväntning--maximering-algoritmen, modifiera den med exponentiell lutning och sedan använda speciellt utvecklade optimeringsfunktioner för att ta hänsyn till prediktionsosäkerheter.Numeriska resultat för konceptuell demonstration erhålls för ett fall av prostatacancer varvid behandlingen levererades med volymetriskt modulerad bågterapi, där det visas att metoderna utvecklade i detta arbete är framgångsrika i att automatiskt generera Paretooptimala planer med tillfredsställande kvalitet och variation medan kliniskt irrelevanta dosregioner utesluts. I fallet då historiska planer används som förkunskap är beräkningstiderna markant kortare än för konventionell MCO.

In the thesis, a calibration method for positions of each accelerometer in an Inertial Sensor Array (IMU) sensor array is designed and implemented. In order to model the motion of the sensor array in the real world, we build up a state space model. Based on the model we use, the problem is to estimate the parameters within the state space model. In this thesis, this problem is solved using Maximum Likelihood (ML) framework and two methods are implemented and analyzed. One is based on Expectation Maximization (EM) and the other is to optimize the cost function directly using Gradient Descent (GD). In the EM algorithm, an ill-conditioned problem exists in the M step, which degrades the performance of the algorithm especially when the initial error is small, and the final Mean Square Error (MSE) curve will diverge in this case. The EM algorithm with enough data samples works well when the initial error is large. In the Gradient Descent method, a reformulation of the problem avoids the ill-conditioned problem. After the parameter estimation part, we analyze the MSE curve of these parameters through the Monte Carlo Simulation. The final MSE curves show that the Gradient Descent based method is more robust in handling the numerical issues of the parameter estimation. The Gradient Descent method is also robust to noise level based on the simulation result. / I denna rapport utvecklas och implementeras en kalibreringsmethod för att skatta positionen för en grupp av accelerometrar placerade i en så kallad IMU sensor array. För att beskriva rörelsen för hela sensorgruppen, härleds en dynamisk tillståndsmodell. Problemställningen är då att skatta parametrarna i tillståndsmodellen. Detta löses med hjälp av Maximum Likelihood-metoden (ML) där två stycken algoritmer implementeras och analyseras. En baseras på Expectation Maximization (EM) och i den andra optimeras kostnadsfunktionen direkt med gradientsökning. I EM-algoritmen uppstår ett illa konditionerat delproblem i M-steget, vilket försämrar algoritmens prestanda, speciellt när det initiala felet är litet. Den resulterande MSE-kurvan kommer att avvika i detta fall. Däremot fungerar EM-algoritmen väl när antalet datasampel är tillräckligt och det initiala felet är större. I gradientsökningsmetoden undviks konditioneringsproblemen med hjälp av en omformulering. Slutligen analyseras medelkvadratfelet (MSE) för parameterskattningarna med hjälp av Monte Carlo-simulering. De resulterande MSE-kurvorna visar att gradientsökningsmetoden är mer robust mot numeriska problem, speciellt när det initiala felet är litet. Simuleringarna visar även att gradientsökning är robust mot brus.

Nous y introduisons une nouvelle classe de distributions bivariées de type Marshall-Olkin, la distribution Erlang bivariée. La transformée de Laplace, les moments et les densités conditionnelles y sont obtenus. Les applications potentielles en assurance-vie et en finance sont prises en considération. Les estimateurs du maximum de vraisemblance des paramètres sont calculés par l'algorithme Espérance-Maximisation. Ensuite, notre projet de recherche est consacré à l'étude des processus de risque multivariés, qui peuvent être utiles dans l'étude des problèmes de la ruine des compagnies d'assurance avec des classes dépendantes. Nous appliquons les résultats de la théorie des processus de Markov déterministes par morceaux afin d'obtenir les martingales exponentielles, nécessaires pour établir des bornes supérieures calculables pour la probabilité de ruine, dont les expressions sont intraitables. / In this contribution, we introduce a new class of bivariate distributions of Marshall-Olkin type, called bivariate Erlang distributions. The Laplace transform, product moments and conditional densities are derived. Potential applications of bivariate Erlang distributions in life insurance and finance are considered. Further, our research project is devoted to the study of multivariate risk processes, which may be useful in analyzing ruin problems for insurance companies with a portfolio of dependent classes of business. We apply results from the theory of piecewise deterministic Markov processes in order to derive exponential martingales needed to establish computable upper bounds of the ruin probabilities, as their exact expressions are intractable.

Cette thèse traite de l’identification de systèmes dynamiques à partir de données échantillonnées à pas variable. Ce type de données est souvent rencontré dans les domaines biomédical, environnemental, dans le cas des systèmes mécaniques où un échantillonnage angulaire est réalisé ou lorsque les données transitent sur un réseau. L’identification directe de modèles à temps continu est l’approche à privilégier lorsque les données disponibles sont échantillonnées à pas variable ; les paramètres des modèles à temps discret étant dépendants de la période d’échantillonnage. Dans une première partie, un estimateur optimal de type variable instrumentale est développé pour estimer les paramètres d’un modèle Box-Jenkins à temps continu. Ce dernier est itératif et présente l’avantage de fournir des estimées non biaisées lorsque le bruit de mesure est coloré et sa convergence est peu sensible au choix du vecteur de paramètres initial. Une difficulté majeure dans le cas où les données sont échantillonnées à pas variable concerne l’estimation de modèles de bruit de type AR et ARMA à temps continu (CAR et CARMA). Plusieurs estimateurs pour les modèles CAR et CARMA s’appuyant sur l’algorithme Espérance-Maximisation (EM) sont développés puis inclus dans l’estimateur complet de variable instrumentale optimale. Une version étendue au cas de l’identification en boucle fermée est également développée. Dans la deuxième partie de la thèse, un estimateur robuste pour l'identification de systèmes à retard est proposé. Cette classe de systèmes est très largement rencontrée en pratique et les méthodes disponibles ne peuvent pas traiter le cas de données échantillonnées à pas variable. Le retard n’est pas contraint à être un multiple de la période d’échantillonnage, contrairement à l’hypothèse traditionnelle dans le cas de modèles à temps discret. L’estimateur développé est de type bootstrap et combine la méthode de variable instrumentale itérative pour les paramètres de la fonction de transfert avec un algorithme numérique de type gradient pour estimer le retard. Un filtrage de type passe-bas est introduit pour élargir la région de convergence pour l’estimation du retard. Tous les estimateurs proposés sont inclus dans la boîte à outils logicielle CONTSID pour Matlab et sont évalués à l’aide de simulation de Monte-Carlo / The output of a system is always corrupted by additive noise, therefore it is more practical to develop estimation algorithms that are capable of handling noisy data. The effect of white additive noise has been widely studied, while a colored additive noise attracts less attention, especially for a continuous-time (CT) noise. Sampling issues of CT stochastic processes are reviewed in this thesis, several sampling schemes are presented. Estimation of a CT stochastic process is studied. An expectation-maximization-based (EM) method to CT autoregressive/autoregressive moving average model is developed, which gives accurate estimation over a large range of sampling interval. Estimation of CT Box-Jenkins models is also considered in this thesis, in which the noise part is modeled to improve the performance of plant model estimation. The proposed method for CT Box-Jenkins model identification is in a two-step and iterative framework. Two-step means the plant and noise models are estimated in a separate and alternate way, where in estimating each of them, the other is assumed to be fixed. More specifically, the plant is estimated by refined instrumental variable (RIV) method while the noise is estimated by EM algorithm. Iterative means that the proposed method repeats the estimation procedure several times until a optimal estimate is found. Many practical systems have inherent time-delay. The problem of identifying delayed systems are of great importance for analysis, prediction or control design. The presence of a unknown time-delay greatly complicates the parameter estimation problem, essentially because the model are not linear with respect to the time-delay. An approach to continuous-time model identification of time-delay systems, combining a numerical search algorithm for the delay with the RIV method for the dynamic has been developed in this thesis. In the proposed algorithm, the system parameters and time-delay are estimated reciprocally in a bootstrap manner. The time-delay is estimated by an adaptive gradient-based method, whereas the system parameters are estimated by the RIV method. Since numerical method is used in this algorithm, the bootstrap method is likely to converge to local optima, therefore a low-pass filter has been used to enlarge the convergence region for the time-delay. The performance of the proposed algorithms are evaluated by numerical examples

This thesis addresses a challenging problem: “how to let the visually impaired ‘see’ others emotions”. We, human beings, are heavily dependent on facial expressions to express ourselves. A smile shows that the person you are talking to is pleased, amused, relieved etc. People use emotional information from facial expressions to switch between conversation topics and to determine attitudes of individuals. Missing emotional information from facial expressions and head gestures makes the visually impaired extremely difficult to interact with others in social events. To enhance the visually impaired’s social interactive ability, in this thesis we have been working on the scientific topic of ‘expressing human emotions through vibrotactile patterns’. It is quite challenging to deliver human emotions through touch since our touch channel is very limited. We first investigated how to render emotions through a vibrator. We developed a real time “lipless” tracking system to extract dynamic emotions from the mouth and employed mobile phones as a platform for the visually impaired to perceive primary emotion types. Later on, we extended the system to render more general dynamic media signals: for example, render live football games through vibration in the mobile for improving mobile user communication and entertainment experience. To display more natural emotions (i.e. emotion type plus emotion intensity), we developed the technology to enable the visually impaired to directly interpret human emotions. This was achieved by use of machine vision techniques and vibrotactile display. The display is comprised of a ‘vibration actuators matrix’ mounted on the back of a chair and the actuators are sequentially activated to provide dynamic emotional information. The research focus has been on finding a global, analytical, and semantic representation for facial expressions to replace state of the art facial action coding systems (FACS) approach. We proposed to use the manifold of facial expressions to characterize dynamic emotions. The basic emotional expressions with increasing intensity become curves on the manifold extended from the center. The blends of emotions lie between those curves, which could be defined analytically by the positions of the main curves. The manifold is the “Braille Code” of emotions. The developed methodology and technology has been extended for building assistive wheelchair systems to aid a specific group of disabled people, cerebral palsy or stroke patients (i.e. lacking fine motor control skills), who don’t have ability to access and control the wheelchair with conventional means, such as joystick or chin stick. The solution is to extract the manifold of the head or the tongue gestures for controlling the wheelchair. The manifold is rendered by a 2D vibration array to provide user of the wheelchair with action information from gestures and system status information, which is very important in enhancing usability of such an assistive system. Current research work not only provides a foundation stone for vibrotactile rendering system based on object localization but also a concrete step to a new dimension of human-machine interaction. / Taktil Video