National Repository of Grey Literature
Not a member yet
574846 research outputs found
Sort by
Wireless EMG sensor
No full textTato práce shromažďuje informace potřebné pro vytvoření kompaktního, bezdrátového a uživatelsky přívětivého elektromyografického (EMG) senzoru. Zabývá se miniaturizací zařízení, bezdrátovým přenosem dat, integrovanými bateriovými systémy, sběrem EMG signálu a možnými případy použití takového senzoru. Jsou diskutovány běžné metody přenosu dat, přičemž jsou zmíněny významné integrované obvody a návrhy obvodů. Praktická část popisuje funkční návrh takového senzoru a vysvětluje jeho klíčové funkce. Kromě elektrického návrhu se značná část zabývá softwarovým návrhem zařízení, návrhem přijímače, navrženým komunikačním protokolem a jeho implikacemi na softwarové rozhraní. Pro vyhodnocení přesnosti, použitelnosti a adekvátnosti senzoru byly změřeny základní amplitudové a frekvenční parametry
Data analysis from Finapres NOVA
No full textThis bachelors thesis focuses on the measurement and subsequent analysis of data obtained using the Finapres NOVA and a smartphone. The theoretical section describes the parameters and capabilities of the Finapres NOVA as well as the cardiovascular parameters it measures. The practical part of the thesis involves measuring PPG, ECG, and PCG signals during various activities and creating a peak detector for PPG using the derivative method. Heart rate was calculated from the detected peaks and compared to a reference. The heart rate, which was calculated from peaks in ECG detected using the provided BUT_ECG_QRSdet_LML detector that combines three QRS complex detection methods, was used as the reference. The results are interpreted as the mean absolute error (MAE). For PPG measured using Finapres NOVA, the MAE is 0.53 beats/min, and for experimental signals extracted from smartphone video, the MAE is 1.62 beats/min for the red channel and 1.87 beats/min for the green channel. In the next part of the thesis, the variability of three Valsalva maneuver trials was analyzed for each subject using cross-correlation. The highest similarity was found between the second and third maneuvers, with an average correlation coefficient of 0.818. Additionally, blood pressure and heart rate values before and after each Valsalva maneuver were compared and statistically evaluated using paired t-tests and Wilcoxon signed-rank tests, depending on the normality of the data. A statistically significant change was observed in the mean systolic blood pressure (p = 0.02162), maximum systolic blood pressure (p = 0,04669) and heart rate (p = 0.00037) after the first maneuver, in the range of diastolic blood pressure (p = 0.00430) and heart rate (p = 0.00025) after the second maneuver and in the range of diastolic blood pressure (p = 0.00165), maximum diastolic blood pressure (p = 0,01289) and heart rate (p = 0.00000) after the third maneuver. Finally, the results of all analyses are discussed
Lens horn antenna for K band
No full textThis bachelor thesis deals with the design of a funnel-shaped conical antenna and a transmit array in the K-band frequency range from 18 to 27 GHz. The aim of this thesis was to design a funnel antenna so that its radiation pattern has the smallest possible angular width of the main lobe and so that side lobes are minimized. This thesis has several main chapters, the first of which is a theoretical part on the radiation of electromagnetic waves and the structure of the antenna, and the main important parameters of the lens and antenna are given. The second chapter deals with the design of the antenna and lens, and the third chapter deals with the design of the final construction and simulations, including the antenna, the transition to coaxial cable, and the lens. The result is a funnel-shaped conical antenna design that meets the requirements and a lens that generally meets the specifications and improves all important antenna parameters described in the first chapter
Proposals for Outdoor Advertising for Generation Z Customers in the Context of Visual Smog
No full textThe thesis focuses on designing outdoor advertising for Oriflame in Brno, aiming to engage women from Generation Z within the context of visual smog. The work is divided into three parts. The first part presents the theoretical background on outdoor advertising, the attractiveness of Generation Z, and visual smog. The second part introduces Oriflame and explores the impact of visual smog on the attractiveness of outdoor advertising – specifically its ability to capture attention and be memorable for women from Generation Z in Brno. The final part of the thesis presents specific outdoor advertising proposals. The findings can serve as a foundation for developing outdoor advertisements that are both attractive and memorable for Generation Z women
Weight Initialization in Neural Networks
No full textTato práce se zabývá technikou obrácené destilace modelu neuronových sítí, při které menší model (učitel) vede počáteční fázi tréninku většího modelu (studenta). Cílem je ověřit, zda tento přístup může sloužit jako efektivní způsob inicializace vah a zlepšit průběh tréninku. Experimenty na klasifikačních úlohách (CIFAR-10 a Fashion-MNIST) ukazují, že destilace po celou dobu tréninku zvyšuje přesnost a odolnost vůči FGSM útokům. U regresní úlohy (California Housing) sice nedošlo ke zlepšení výsledné přesnosti, ale metoda vedla ke snížení testovací ztráty při adversariálním útoku. Výsledky naznačují, že obrácená destilace může podpořit stabilnější a odolnější trénink neuronových sítí
Study of the New Tram Track between Videnska Street and New Main Station in Brno
No full textThe bachelor‘s thesis disscuses about designing tram railway in Brno between new prepared Main train station and Videňska street. The design was made in variants according to different needs for accessibility and drive through the area. One of designed variants was chosen and inspected more in detail. There is included comparsion between variants according to design and needs for designing
Marketing Mix Proposal of the Newly Established Building Company Division
No full textThe thesis focuses on the proposal of a marketing mix for a newly emerging division of a construction company. The theoretical part is devoted to the definition of key concepts and principles necessary for the elaboration of the topic. The analytical part examines the current state of the marketing mix in the company. The proposal part presents specific proposals aimed at reaching and acquiring new customers
Convolutional neural networks for detection of handwritten characters
No full textThe bachelor thesis deals with the problem of handwritten character detection. The theoretical part describes machine and deep learning, the problem of learning from small amounts of data, computer vision and object detection in images. It also includes a brief insight into convolutional neural networks and transformers. The practical part focuses on the design, implementation and subsequent testing of models and approaches for detection of the handwritten character "a". The performance of modern detectors YOLOv8, YOLO11, DINO, Faster R-CNN and custom proposed detector is tested using transfer learning and training on a custom created dataset with data augmentations. The testing also includes segmentation models YOLOv8-seg, YOLO11-seg and Mask R-CNN. Another important part of the solution is testing of sliced inference and creation of a simple web platform for fast and easy annotation of given characters and model training using modified active learning. The results of this work show that YOLO11m is the most suitable model for the problem at hand. Also, the positive effect of sliced inference on the number of detected characters and the suitability of the combination of active learning and self-training for iterative model training is demonstrated
Evolutionary Neural Network Optimization Using a Super Net
No full textThis bachelor's thesis presents a methodology for the evolutionary optimization of convolutional neural networks (CNNs) using the multi-objective genetic algorithm NSGA-II. The goal is to balance classification accuracy and computational efficiency, particularly in terms of inference latency on the target platform, the Hailo-8L accelerator. The proposed approach builds upon a pre-trained supernet, enabling efficient exploration of the architecture space without fully retraining each candidate. To accelerate evaluation, a latency predictor is introduced, which estimates inference time without requiring hardware deployment. Experimental results on the CIFAR-10 and CIFAR-100 datasets demonstrate that the method can produce architectures with favorable accuracy-latency trade-offs. This approach offers an effective path to automated CNN design tailored to specific hardware constraints and lays the groundwork for further research in hardware-aware neural architecture search
Design of Energy-Efficient Family House in Hustopeče
No full textThe Bachelor´s thesis focuses on designing nearly-zero energy detached house located in Hustopeče. It has two main parts: architectural design and building services. It is an L - Shaped detached house of two storeys with a flat roof. The residential area on the first floor includes a wind lobby with a walk-in closet, a corridor leading to a study, a storage, a bathroom, a toilet, and an open-space kitchen. The second floor features a corridor that is connected to a walk-in closet, a master bedroom, two bedrooms, a bathroom, and a roof terrace. The technical section comprises a storage room, a utility room, and a garage. Foundations are concrete strips. The horizontal load - bearing structures are design of RC floor slabs, while vertical structures are built of ceramic blocks. The house has flat green roofs and one is partially walkable. The heat source is an air-to-water heat pump. Individual rooms are heated by floor heating. Bathrooms and toilets are equipped with towel radiators. The utility rooms are heated by radiators. The building is ventilated by mechanical ventilation with ceiling - suspended heat exchanger in the utility room. Fresh air is supplied to the living rooms and is exhausted from the bathrooms through the corridor and technical facilities. Drinking water is supplied to the building through a new municipal water main. Water heating is provided by an in-built heating tank into a 190l heat pump. The sewage is discharged through a new connection to the sewer. Rainwater from the roofs is drained into a storage tank in the garden. The water is reused for flushing and watering. Excess water is discharged into an infiltration system. There is a photovoltaic system on the roof over the second floor