O aumento constante da demanda por sistemas computacionais cada vez mais eficientes tem motivado a busca por sistemas híbridos customizados compostos por GPP (General Purpose Processor), FPGAs (Field-Programmable Gate Array) e GPUs (Graphics Processing Units). Quando utilizados em conjunto possibilitam otimizar a relação entre desempenho e consumo de energia. Tais sistemas dependem de técnicas que façam o mapeamento mais adequado considerando o perfil do código fonte. Nesse sentido, este projeto propõe uma técnica para realizar o mapeamento entre GPP e FPGA. Para isso, utilizou-se como base uma abordagem de mineração de dados que avalia a similaridade entre código fonte. A técnica aqui desenvolvida obteve taxas de acertos de 65,67% para códigos sintetizados para FPGA com a ferramenta LegUP e 59,19% para Impulse C, considerando que para GPP o código foi compilado com o GCC (GNU Compiler Collection) utilizando o suporte a OpenMP. Os resultados demonstraram que esta abordagem pode ser empregada como um ponto de decisão inicial no processo de mapeamento em sistemas híbridos, somente analisando o perfil do código fonte sem que haja a necessidade de execução do mesmo para a tomada de decisão. / The steady increasing on demand for efficient computer systems has been motivated the search for customized hybrid systems composed by GPP (general purpose processors), FPGAs (Field- Programmable Gate Array) and GPUs (Graphics Processing Units). When they are used together allow to exploit their computing resources to optimize performance and power consumption. Such systems rely on techniques make the most appropriate mapping considering the profile of source code. Thus, this project proposes a technique to perform the mapping between GPP and FPGA. For this, it is applied a technique based on a data mining approach that evaluates the similarity between source code. The proposed method obtained hit rate 65.67% for codes synthesized in FPGA using LegUP tool and 59.19% for Impulse C tool, whereas for GPP, the source code was compiled on GCC (GNU Compiler Collection) using OpenMP. The results demonstrated that this approach can be used as an initial decision point on the mapping process in hybrid systems, only analyzing the profile of the source code without the need for implementing it for decision-making.

The thesis deals with a concept and creation of classifiers of defects found on continuous production lines. The first part presents an overview of methods used for image classification and a analysis of defects. The main part of the thesis consist of a description of created classifier interface and graphical user interface for classifier. The last part sums up reliability of each implemented classifer.

This work deals with speaker´s age and gender recognition. At the beginning it introduces the practical usage of this application and discusses the solutions available. The theoretical part of the thesis specifies the feature extraction and reduction methods and speech databases used in the experiments. The practical part describes the recognizer implemented in the Emotional tool and in two chapters describes the individual experiments. Regarding speaker´s gender estimation; we focused on the impact of the emotional state and speaker's age on the classification process. The two remain experiments were dedicated for general gender estimation performed by using two different classifiers – GMM and k-NN. These two classifiers were used in age estimation as well. In this case, four Group of age was formed and two different feature sets namely: segmental and suprasegmental were exploited four groups

This dissertation deals with analysis of current side canal by means of neural network. First part describes basis of cryptografy and dilemma of side canal. In the second part is theoretickly described neural network and correlative analysis. Third part describes practical analysis of calibres of current side canals by means of classifier which uses neural network in Matlab surrounding. This classifier is confronted with classifier which uses correlative analysis.

This master‘s thesis describes the identification faces in real terms. It includes an overview of current methods of detection faces by the classifiers. It also includes various methods for detecting faces. The second part is a description of two programs designed to identify persons. The first program operates in real time under laboratory conditions, where using web camera acquires images of user's face. This program is designed to speed recognition of persons. The second program has been working on static images, in real terms. The main essence of this method is successful recognition of persons, therefore the emphasis on computational complexity. The programs I used a staged method of PCA, LDA and kernel PCA (KPCA). The first program only works with the PCA method, which has good results with respect to the success and speed of recognition. In the second program to compare methods, which passed the best method for KPCA.

Speech processing systems have been developed for many years but the integration into devices had started with the deployment of the modern powerful computational systems. This dissertation thesis deals with development of the keyword detection system in speech data. The proposed detection system is based on the Large Margin and Kernel methods and the key part of the system is phoneme classifier. Two hierarchical frame-based classifiers have been proposed -- linear and non-linear. An efficient training algorithm for each of the proposed classifier have been introduced. Simultaneously, classifier based on the Gaussian Mixture Models with the implementation of the hierarchical structure have been proposed. An important part of the detection system is feature extraction and therefor all algorithms were evaluated on the current most common feature techniques. A part of the thesis technical solution was implementation of the keyword detection system in MATLAB and design of the hierarchical phoneme structure for Czech language. All of the proposed algorithms were evaluated for Czech and English language over the DBRS and TIMIT speech corpus.

This thesis deals with the procedural generation of building models based on a given pattern. The community project OpenStreetMap is used for obtaining datasets that create the buildings platform patterns. A brief survey of classifiers and formal grammars for modeling is introduced. Designing an estate classifier and algorithm for building generation is practical aspect of this thesis, including the algorithm implementation. 3D output meshes are rendered using OpenGL in real-time.

A biometric method is a more secure way of personal identification than passwords. This thesis examines the iris as a personal identifier with the use of neural networks as the classifier. A comparison of different feature extraction methods that include the Fourier transform, discrete cosine transform, the eigen analysis method, and the wavelet transform, is performed. The robustness of each method, with respect to distortion and noise, is also studied.

In recent years wearable devices have been advancing at a rapid pace and one of the largest growing segments is the smart glass segment. In this thesis the feasibility of today’s ARM-based smart glasses are evaluated for automatic license plate recognition (ALPR). The license plate is by far the most prominent visual feature to identify a spe- cific vehicle, and exists on both old and newly produced vehicles. This thesis propose an ALPR system based on a sequence of vertical edge detection, a cascade classifier, verti- cal and horizontal projection as well as a general purpose optical character recognition library. The study further concludes that the optimal input resolution for license plate detection using vertical edges is 640x360 pixels and that the license plate need to be at least 20 pixels high or the characters 15 pixels high in order to successfully segment the plate and recognize each character. The separate stages were successfully implemented into a complete ALPR system that achieved 79.5% success rate while processing roughly 3 frames per second when running on a pair of Google Glass. / Under de senaste åren har området wearables avancerat i snabb takt, och ett av de snabbast växande segmenten är smarta glaögon. I denna examensuppsats utvärderas lämpligheten av dagens ARM-baserade smarta glasögon med avseende på automatisk registreringsskyltigenkänning. Registreringsskylten är den i särklass mest framträdande visuella egenskapen som kan användas för att identifiera ett specifikt fordon, och den finns på både gamla och nyproducerade fordon. Detta examensarbete föreslår ett system för automatisk registreringsskyltigenkänning baserat på en följd av vertikal kantdetektering, en kaskad av boostade klassificerare, vertikal och horisontell projektion samt ett optiskt teckenigenkänningsbibliotek. Studien konstaterar vidare att den optimala upplösningen för registreringsskyltdetektion med hjälp av vertikala kanter på smarta glasögonär 640x360 pixlar och att registreringsskylten måste vara minst 20 pixlar hög eller tecknen 15 pixlar höga för att registreringsskylten framgångsrikt skall kunna segmenteras samt tecken identifieras. De separata stegen implementerades framgångsrikt till ett system för automatisk registreringsskyltigenkänning på ett par Google Glass och lyckades känna igen 79,5% av de testade registreringsskyltarna, med en hastighet av ungefär 3 bilder per sekund.

Data imbalance has been a challenge in many areas of automatic classification. Many popular approaches including over-sampling, under-sampling, and Synthetic Minority Oversampling Technique (SMOTE) have been developed and tested in previous research. A big problem with these techniques is that they try to solve the problem by modifying the original data rather than truly overcome the imbalance and let the classifiers learn. For tasks in areas like remote sensing and depression detection, the imbalanced data challenge also exists. Researchers have made efforts to overcome the challenge by adopting methods at the data pre-processing step. However, in remote sensing and depression detection tasks, the main interest is still on applying different new classifiers such as deep learning which has powerful classification ability but still do not consider data imbalance as prime factor of lower classification performance. In this thesis, we demonstrate the performance of K-CR in our evaluation experiments on a urban land cover classification dataset and on two depression detection datasets. The latter two datasets consist in social media texts (tweets), therefore we propose to adopt a feature selection technique Term Frequency - Category-Based Term Weights (TF-CBTW) and various word embedding techniques (Word2Vec, FastText, GloVe, and language model BERT). This feature selection method was not applied before in similar settings and we show that it helps to improve the efficiency and the results of the K-CR classifier. Our three experiments show that K-CR can achieve comparable performance on the majority classes and better performance on minority classes when compared to other classifiers such as Random Forest, K-Nearest Neighbour, Support Vector Machines, Multi-layer Perception, Convolutional Neural Networks, and Long Short-Term Memory.