Dans la première partie de cette thèse, nous cherchons à estimer l'intensité d'un processus de comptage par des techniques d'apprentissage statistique en grande dimension. Nous introduisons une procédure d'estimation basée sur la pénalisation par variation totale avec poids. Un premier ensemble de résultats vise à étudier l'intensité sous une hypothèse a priori de segmentation sparse. Dans une seconde partie, nous étudions la technique de binarisation de variables explicatives continues, pour laquelle nous construisons une régularisation spécifique à ce problème. Cette régularisation est intitulée ``binarsity'', elle pénalise les valeurs différentes d'un vecteur de paramètres. Dans la troisième partie, nous nous intéressons à la régression dynamique pour les modèles d'Aalen et de Cox avec coefficients et covariables en grande dimension, et pouvant dépendre du temps. Pour chacune des procédures d'estimation proposées, nous démontrons des inégalités oracles non-asymptotiques en prédiction. Nous utilisons enfin des algorithmes proximaux pour résoudre les problèmes convexes sous-jacents, et nous illustrons nos méthodes sur des données simulées et réelles. / In the first part of this thesis, we deal with the problem of learning the inhomogeneous intensity of a counting process, under a sparse segmentation assumption. We introduce a weighted total-variation penalization, using data-driven weights that correctly scale the penalization along the observation interval. In the second part, we study the binarization technique of continuous features, for which we construct a specific regularization. This regularization is called “binarsity”, it computes the different values of a parameter. In the third part, we are interested in the dynamic regression models of Aalen and Cox with time-varying covariates and coefficients in high-dimensional settings. For each proposed estimation procedure, we give theoretical guaranties by proving non-asymptotic oracle inequalities in prediction. We finally present proximal algorithms to solve the underlying studied convex problems, and we illustrate our methods with simulated and real datasets.

This diploma thesis deal with safety components of trailer and semitrailer technology, especially the elements for underrun protection. The first part deals with the legislative requirements on which safety elements are based. In another part are mentioned the construction solutions of underrun protection devices from the portfolio of company SVAN Chrudim s.r.o. Finally, the computational part is processed to verify sufficient strength of underpass protections.

This diploma work deals with music genre recognition using the techniques of Music Information Retrieval. It contains a brief description of the principle of this research area and its subfield called Music Genre Recognition. The following chapter includes selection of the most suitable parameters for describing music genres. This work further characterizes machine learning methods used in this field of research. The next chapter deals with the descriptions of music datasets created for genre classification studies. Subsequently, there is a draft and evaluation of the system for music genre recognition. The last part of this work describes the results of partial parameter analysis, dependence of genre classification accuracy on the amount of parameters and contains a discussion on the causes of classification accurancy for the individual genres.

The aim of this proposal is socialogical analysis, interpretation, construction and urbanistic recommendation related to an impact of social changes in specific neighborhood on life standarts and real estates‘ costs in the area. Social changes could be negative, such as social degradation of certain area, worsening of quality of the environment, worsening relationships and also social deviation (for example criminality). Positive changes represents mostly gentrification, which includes positive changes of social and tangible environment. These are followed by price encrease of life standarts which is displacing old settlers. In our experience these areas of life standarts and real estates prices are linked and it is possible to use synergistic influence of changes to adjust tangible and social environment.

This thesis is focused on texture analysis methods. The project contains an overview of widely used methods. The main aim of the thesis is to develop a method for texture analysis of retinal images, which will be used for distinction of two patient groups, one with glaucoma eyes and one healthy. It is observed that glaucoma patients don´t have a texture on the eye ground. Preprocessing of the images is found by transfer of the image to different color spaces to achieve the best emphasis of the eye ground texture. Co-occurrence matrix is chosen for texture analysis of this data. The thesis contains detail description of the chosen solutions and feature discussion and the result is a list of features, which can be used for distinction between glaucoma and healthy eyes. The method is implemented in Matlab environment.

The master thesis deals with the topic of detecting faces in digital images. There are generally described and classified the most frequently used methods and discussed their advantages and disadvantages. More detailed is described skin color detection, eye and mouth detection and are teoretically described machine learning algorithms and detection based on Haar-classifiers. The work aims to implementation of these methods in the OpenCV library, it refers to practical application of them a finally compares different provided trained files.

This document describes the three methods for the detection and classification of paralinguistic expressions such as laughing and crying from usual speech by analysis of the audio signal. The database of records was originally designed for this purpose. When analyzing everyday dialogs, music might be included, so the database was extended by four new classes as speech, music, singing with music and usual speech with background music. Feature extraction, feature reduction and classification are common steps in recognizing for all three methods. Difference of the methods is given by classification process in detail. One classification of all six classes at once is proposed in the first method called straight approach. In the second method called decision tree oriented approach we are using five intuitive sub classifiers in the tree structure and the final method uses for classification emotion coupling approach. The best features were reduced by feature evaluation using F-ratio and GMM classifiers were used for the each classification part.

The main theme of the thesis is to examine the influence of alcohol on the speech apparatus and speech signal. The first part is focused on symptoms and detection of alcohol concentration in the human body. The following part describes somescientific publications and projects, which dealt witha a similar theme. Also the czech documentation to german database ALC was created. Based on phonetic knowledge, Czech text was compiled. Different speakers were reading this text so we go tour own database of alcoholic and sober speech. Samples from individual speakers are processed using linear prediction, formant and cepstral analysis in MATLAB and the effect of alcohol on selected parameters of speech signal is evaluated.

The presented master thesis deals with the evaluation of the position of the iris compared with surroundings of the eye. This technique is supposed to be use for computer control. The created software works in real-time mode, the pictures are taken with an ordinary webcam. The first part of the work presents basic algorithms used in computer vision for edit of images. The following part is focused on abilities of methods to find the face, eyes and to detect the iris. The detection and the subsequent separation of the face is based on the recognition of skin colour in the YCbCr color space. The position of eye is then searched in the face by Haar-like features. The darkest part of the eye is found by horizontal projection from surroundings and the seed point is started from this place. From the area which is filled by the seed method (Flood Fill) and which shows the iris, the cursor movement is controlled by obtained x and y position.

Changes in chromosome number and structure may cause serious diseases. Cytogenetic tests leadin to set of karyotype are done for detecting these abnormalities. Chromosomes are visualised with proper methods and karyotype is made up most often. Manual karyotyping is time-consuming and expensive task. Because of this, researchers have been developing automated karyotyping systems. Karyotyping systems classify chromosomes into classes based on their characteristic features. Overlapping and bent chromosomes are limitations for automatic classification since they ocur at almost every mitosis. Accuracy and reliability of karyotyping systems still depend on the human intervention. Overcoming of these problems and development of fully automated system is the aim of modern approaches.