Main purpose of the thesis is to study the processes and methods of isolated words recognition. In the theoretical part a basic principals are explained. The practical part is about the program creating using these principles in practice. For isolated words recognition Hidden Markov Models (HMM) are used, for obtaining decision symptoms cepstral analysis is chosen.

This masters thesis describes possibilities and principles of gesture-based computer interface. The work describes general approaches for gesture control.  It also deals with implementation of the selected detection method of the hands and fingers using depth maps loaded form Kinect sensor. The implementation also deals with gesture recognition using hidden Markov models. For demonstration purposes there is also described implementation of a simple photo viewer that uses developed gesture-based computer interface. The work also focuses on quality testing and accuracy evaluation for selected gesture recognizer.

<p>U disertaciji su opisani postupci sinteze ekspresivnog govora<br />korišćenjem parametarskih pristupa. Pokazano je da se korišćenjem<br />dubokih neuronskih mreža dobijaju bolji rezultati nego korišćenjem<br />skrivenix Markovljevih modela. Predložene su tri nove metode za<br />sintezu ekspresivnog govora korišćenjem dubokih neuronskih mreža:<br />metoda kodova stila, metoda dodatne obuke mreže i arhitektura<br />zasnovana na deljenim skrivenim slojevima. Pokazano je da se najbolji<br />rezultati dobijaju korišćenjem metode kodova stila. Takođe je<br />predložana i nova metoda za transplantaciju emocija/stilova<br />bazirana na deljenim skrivenim slojevima. Predložena metoda<br />ocenjena je bolje od referentne metode iz literature.</p> / <p>In this thesis methods for expressive speech synthesis using parametric<br />approaches are presented. It is shown that better results are achived with<br />usage of deep neural networks compared to synthesis based on hidden<br />Markov models. Three new methods for synthesis of expresive speech using<br />deep neural networks are presented: style codes, model re-training and<br />shared hidden layer architecture. It is shown that best results are achived by<br />using style code method. The new method for style transplantation based on<br />shared hidden layer architecture is also proposed. It is shown that this<br />method outperforms referent method from literature.</p>

>Magister Scientiae - MSc / This thesis presents a system for performing whole gesture recognition for South African Sign Language. The system uses feature vectors combined with Hidden Markov models. In order to construct a feature vector, dynamic segmentation must occur to extract the signer's hand movements. Techniques and methods for normalising variations that occur when recording a signer performing a gesture, are investigated. The system has a classification rate of 69%.

La production d'énergie renouvelable et la consommation d'électricité dépendent largement des conditions météorologiques : température, précipitations, vent, rayonnement solaire... Ainsi, pour réaliser des études d'impact sur l'équilibre offre-demande, on peut utiliser un générateur de temps, c'est-à-dire un modèle permettant de simuler rapidement de longues séries de variables météorologiques réalistes, au pas de temps journalier. L'une des approches possibles pour atteindre cet objectif utilise les modèles de Markov caché : l'évolution des variables à modéliser est supposée dépendre d'une variable latente que l'on peut interpréter comme un type de temps. En adoptant cette approche, nous proposons dans cette thèse un modèle permettant de simuler simultanément la température, la vitesse du vent et les précipitations, en tenant compte des non-stationnarités qui caractérisent les variables météorologiques. D'autre part, nous nous intéressons à certaines propriétés théoriques des modèles de Markov caché cyclo-stationnaires : nous donnons des conditions simples pour assurer leur identifiabilité et la consistance forte de l'estimateur du maximum de vraisemblance. On montre aussi cette propriété de l'EMV pour des modèles de Markov caché incluant des tendances de long terme sous forme polynomiale. / Renewable energy production and electricity consumption both depend heavily on weather: temperature, precipitations, wind, solar radiation... Thus, making impact studies on the supply/demand equilibrium may require a weather generator, that is a model capable of quickly simulating long, realistic times series of weather variables, at the daily time step. To this aim, one of the possible approaches is using hidden Markov models : we assume that the evolution of the weather variables are governed by a latent variable that can be interpreted as a weather type. Using this approach, we propose a model able to simulate simultaneously temperature, wind speed and precipitations, accounting for the specific non-stationarities of weather variables. Besides, we study some theoretical properties of cyclo-stationary hidden Markov models : we provide simple conditions of identifiability and we show the strong consistency of the maximum likelihood estimator. We also show this property of the MLE for hidden Markov models including long-term polynomial trends.

This master's thesis deals with design of speech recognition algorithms with consideration of target technology, which is platform combinating digital signal processing and field programmable gate array. Algorithms for speech recognition includes: feature extraction of Melfrequency cepstral coefficients, hidden Markov models and their evaluation by Viterbi algorithm.

This thesis focuses on comparing an online parameter estimator to an offline estimator, both based on the PaRIS-algorithm, when estimating parameter values for a stochastic volatility model. By modeling the stochastic volatility model as a hidden Markov model, estimators based on particle filters can be implemented in order to estimate the unknown parameters of the model. The results from this thesis implies that the proposed online estimator could be considered as a superior method to the offline counterpart. The results are however somewhat inconclusive, and further research regarding the subject is recommended. / Detta examensarbetefokuserar på att jämföra en online och offline parameter-skattare i stokastiskavolatilitets modeller. De två parameter-skattarna som jämförs är båda baseradepå PaRIS-algoritmen. Genom att modellera en stokastisk volatilitets-model somen dold Markov kedja, kunde partikelbaserade parameter-skattare användas föratt uppskatta de okända parametrarna i modellen. Resultaten presenterade idetta examensarbete tyder på att online-implementationen av PaRIS-algorimen kanses som det bästa alternativet, jämfört med offline-implementationen.Resultaten är dock inte helt övertygande, och ytterligare forskning inomområdet

Image Captioning (IC) aims to generate a coherent and comprehensive textual description that summarizes the complex content of an image. It is a combination of computer vision and natural language processing techniques to encode the visual features of an image and translate them into a sentence. In the context of remote sensing (RS) analysis, IC has been emerging as a new research area of high interest since it not only recognizes the objects within an image but also describes their attributes and relationships. In this thesis, we propose several IC methods for RS image analysis. We focus on the design of different approaches that take into consideration the peculiarity of RS images (e.g. spectral, temporal and spatial properties) and study the benefits of IC in challenging RS applications. In particular, we focus our attention on developing a new decoder which is based on support vector machines. Compared to the traditional decoders that are based on deep learning, the proposed decoder is particularly interesting for those situations in which only a few training samples are available to alleviate the problem of overfitting. The peculiarity of the proposed decoder is its simplicity and efficiency. It is composed of only one hyperparameter, does not require expensive power units and is very fast in terms of training and testing time making it suitable for real life applications. Despite the efforts made in developing reliable and accurate IC systems, the task is far for being solved. The generated descriptions are affected by several errors related to the attributes and the objects present in an RS scene. Once an error occurs, it is propagated through the recurrent layers of the decoders leading to inaccurate descriptions. To cope with this issue, we propose two post-processing techniques with the aim of improving the generated sentences by detecting and correcting the potential errors. They are based on Hidden Markov Model and Viterbi algorithm. The former aims to generate a set of possible states while the latter aims at finding the optimal sequence of states. The proposed post-processing techniques can be injected to any IC system at test time to improve the quality of the generated sentences. While all the captioning systems developed in the RS community are devoted to single and RGB images, we propose two captioning systems that can be applied to multitemporal and multispectral RS images. The proposed captioning systems are able at describing the changes occurred in a given geographical through time. We refer to this new paradigm of analysing multitemporal and multispectral images as change captioning (CC). To test the proposed CC systems, we construct two novel datasets composed of bitemporal RS images. The first one is composed of very high-resolution RGB images while the second one of medium resolution multispectral satellite images. To advance the task of CC, the constructed datasets are publically available in the following link: https://disi.unitn.it/~melgani/datasets.html. Finally, we analyse the potential of IC for content based image retrieval (CBIR) and show its applicability and advantages compared to the traditional techniques. Specifically, we focus our attention on developing a CBIR systems that represents an image with generated descriptions and uses sentence similarity to search and retrieve relevant RS images. Compare to traditional CBIR systems, the proposed system is able to search and retrieve images using either an image or a sentence as a query making it more comfortable for the end-users. The achieved results show the promising potentialities of our proposed methods compared to the baselines and state-of-the art methods.

The hidden Markov model (HMM) is one of the workhorse tools in, for example, statistical signal processing and machine learning. It has found applications in a vast number of fields, ranging all the way from bioscience to speech recognition to modeling of user interactions in social networks. In an HMM, a latent state transitions according to Markovian dynamics. The state is only observed indirectly via a noisy sensor – that is, it is hidden. This type of model is at the center of this thesis, which in turn touches upon three main themes. Firstly, we consider how the parameters of an HMM can be estimated from data. In particular, we explore how recently proposed methods of moments can be combined with more standard maximum likelihood (ML) estimation procedures. The motivation for this is that, albeit the ML estimate possesses many attractive statistical properties, many ML schemes have to rely on local-search procedures in practice, which are only guaranteed to converge to local stationary points in the likelihood surface – potentially inhibiting them from reaching the ML estimate. By combining the two types of algorithms, the goal is to obtain the benefits of both approaches: the consistency and low computational complexity of the former, and the high statistical efficiency of the latter. The filtering problem – estimating the hidden state of the system from observations – is of fundamental importance in many applications. As a second theme, we consider inverse filtering problems for HMMs. In these problems, the setup is reversed; what information about an HMM-filtering system is exposed by its state estimates? We show that it is possible to reconstruct the specifications of the sensor, as well as the observations that were made, from the filtering system’s posterior distributions of the latent state. This can be seen as a way of reverse engineering such a system, or as using an alternative data source to build a model. Thirdly, we consider Markov decision processes (MDPs) – systems with Markovian dynamics where the parameters can be influenced by the choice of a control input. In particular, we show how it is possible to incorporate prior information regarding monotonic structure of the optimal decision policy so as to accelerate its computation. Subsequently, we consider a real-world application by investigating how these models can be used to model the treatment of abdominal aortic aneurysms (AAAs). Our findings are that the structural properties of the optimal treatment policy are different than those used in clinical practice – in particular, that younger patients could benefit from earlier surgery. This indicates an opportunity for improved care of patients with AAAs. / <p>QC 20180301</p>

Detecting a malicious hacker intruding on a network system can be difficult. This challenge is made even more complex by the network activity generated by normal users and by the fact that it is impossible to know the hacker’s exact actions. Instead, the defender of the network system has to infer the hacker’s actions by statistics collected by the intrusion detection system, IDS. This thesis investigates the performance of hidden Markov models, HMM, to detect an intrusion automatically under different background activities generated by normal users. Furthermore, background subtraction techniques with inspiration from computer vision are investigated to see if normal users’ activity can be filtered out to improve the performance of the HMMs.The results suggest that the performance of HMMs are not sensitive to the type of background activity but rather to the number of normal users present. Furthermore, background subtraction enhances the performance of HMMs slightly. However, further investigations into how background subtraction performs when there are many normal users must be done before any definitive conclusions. / Det kan vara svårt att upptäcka en hackare som gör intrång i ett nätverkssystem. Utmaningen blir ännu större genom nätverksaktiviteten som genereras av vanliga användare och av det faktum att det är omöjligt att veta hackarens exakta handlingar. Istället måste nätverkssystemets försvarare använda insamlad data från intrångsdetekteringssystemet, IDS, för att estimera hackarens handlingar. Detta arbete undersöker förmågan hos dolda Markovmodeller, HMM, att automatiskt upptäcka dataintrång under olika typer av bakgrundsaktiviteter som genereras av normala användare. Dessutom undersöks bakgrundssubtraktionstekniker med inspiration från datorseende för att se om normala användares aktivitet kan filtreras bort för att förbättra prestanda hos HMM. Resultaten tyder på att prestandan för HMM inte är känsliga för typen av bakgrundsaktivitet utan snarare för antalet närvarande normala användare. Dessutom förbättrar bakgrundssubtraktion prestandan hos HMM. Det krävs dock mer forskning för att dra definitiva slutsatser kring vilken effekt bakgrundssubstitution har när antalet normala användare är stort.