Prezenční studium

4 roky

doc. Ing. Martin Štumpf, Ph.D.

Předseda :
doc. Ing. Martin Štumpf, Ph.D.
Člen interní :
doc. Ing. Jaroslav Láčík, Ph.D.
doc. Ing. Jiří Petržela, Ph.D.
prof. Ing. Aleš Prokeš, Ph.D.
prof. Dr. Ing. Zbyněk Raida
doc. Ing. Roman Šotner, Ph.D.
doc. Ing. Tomáš Götthans, Ph.D.
Člen externí :
Ing. Ondřej Číp, Ph.D.
doc. Ing. Milan Polívka, Ph.D.

Poskytnout doktorské vzdělání absolventům magisterského vysokoškolského studia v oblasti elektroniky a komunikačních technologií. Prohloubit teoretické znalosti studentů ve vybraných částech vyšší matematiky a fyziky a dát jím též potřebné vědomosti a praktické dovednosti z aplikované informatiky a výpočetní techniky. Naučit je metodám vědecké práce.

Absolvent bude umět řešit vědecké a složité technické úlohy v oblasti elektroniky a elektronických komunikací. Absolventi doktorského studijního programu "Elektronika a komunikační technologie" budou v oblasti elektroniky a sdělovací techniky schopni pracovat jako vědečtí a výzkumní pracovníci v základním či aplikovaném výzkumu, jako specializovaní odborníci vývoje, konstrukce a provozu v různých výzkumných a vývojových institucích, elektrotechnických a elektronických výrobních firmách a společnostech a u uživatelů komunikačních systémů a zařízení, přičemž zde budou schopni tvůrčím způsobem využívat moderní výpočetní komunikační a měřicí techniku.

Absolventi doktorského studijního programu "Elektronika a komunikační technologie" jsou schopni samostatně řešit složité vědecké a technické úlohy v oblasti elektroniky a komunikací. Díky kvalitnímu rozvinutému teoretickému vzdělání a specializaci ve vybraném oboru jsou absolventi doktorského studia vyhledáváni jako specialisté v oblasti elektroniky a komunikační techniky. Absolventi doktorského studijního programu budou schopni pracovat v oblasti elektroniky a sdělovací techniky jako vědečtí a výzkumní pracovníci v základním či aplikovaném výzkumu, jako specializovaní odborníci vývoje, konstrukce a provozu v různých výzkumných a vývojových institucích, elektrotechnických a elektronických výrobních firmách a společnostech, přičemž zde budou schopni tvůrčím způsobem využívat moderní výpočetní komunikační a měřicí techniku.

Studium doktoranda probíhá podle individuálního studijního plánu, který zpracuje v úvodu studia školitel doktoranda ve spolupráci s doktorandem. V individuálním studijním plánu jsou specifikovány všechny povinnosti stanovené v souladu se Studijním a zkušebním řádem VUT, které musí doktorand k úspěšnému ukončení studia splnit. Tyto povinnosti jsou časově rozvrženy do celého období studia, jsou bodově ohodnoceny a v pevně daných termínech probíhá kontrola jejich plnění.
Student si zapíše a vykoná zkoušky z povinných (Návrh moderních elektronických obvodů, Moderní digitální bezdrátová komunikace), minimálně dvou povinně volitelných předmětů ohledem na zaměření jeho disertační práce, a dále minimálně dvou volitelných předmětů (Angličtina pro doktorandy, Řešení inovačních zadání, Vědecké publikování od A do Z)
Ke státní doktorské zkoušce se může student přihlásit až po vykonání všech zkoušek předepsaných jeho individuálním studijním plánem. Před státní doktorskou zkouškou student vypracuje pojednání k disertační práci, v němž detailně popíše cíle práce, důkladné zhodnocení stavu poznání v oblasti řešené disertace, charakteristiku metod, které hodlá při řešení uplatňovat.
Obhajoba pojednání, které je oponováno, je součástí státní doktorské zkoušky. V další části zkoušky musí student prokázat hluboké teoretické i praktické znalosti v oblasti elektrotechniky, elektroniky, komunikační techniky, obecné teorie obvodů a elektromagnetického pole, zpracování signálů, anténní a vysokofrekvenční techniky. Státní doktorská zkouška probíhá ústní formou a kromě diskuze nad pojednáním k disertační práci se také skládá z tematických okruhů týkajících se povinných a povinně volitelných předmětů.
K obhajobě disertační práce se student hlásí po vykonání státní doktorské zkoušky a po splnění podmínek pro ukončení, jakými jsou účast na výuce, vědecká a odborná činnost (tvůrčí činnost), a minimálně měsíční studijní nebo pracovní stáž na zahraniční instituci anebo účast na mezinárodním tvůrčím projektu.

Studium doktoranda probíhá podle individuálního studijního plánu (dále jen ISP), který zpracuje v úvodu studia školitel doktoranda ve spolupráci s doktorandem. Individuální studijní plán je pro doktoranda závazný. Jsou v něm specifikovány všechny povinnosti stanovené v souladu se Studijním a zkušebním řádem VUT, které musí doktorand k úspěšnému ukončení studia splnit. Tyto povinnosti jsou časově rozvrženy do celého období studia, jsou bodově ohodnoceny a v pevně daných termínech probíhá kontrola jejich plnění. Průběžné bodové hodnocení všech aktivit doktoranda je vedeno v dokumentu „Celkové bodové hodnocení doktoranda“ a je součástí ISP. Při zahájení dalšího roku studia pak školitel do ISP zaznamená případné změny. Nejpozději do 15. 10. každého roku studia odevzdává doktorand vytištěný a podepsaný ISP na vědeckém oddělení fakulty ke kontrole a založení.
Během prvních čtyř semestrů skládá doktorand zkoušky z povinných, povinně volitelných anebo volitelných předmětů pro splnění bodových limitů ze Studijní oblasti, a současně se intenzivně zabývá vlastním studiem a analýzou poznatků v oboru stanoveném tématem disertační práce a průběžným publikováním takto získaných poznatků a vlastních výsledků. V dalších semestrech se doktorand již více soustřeďuje na výzkum a vývoj, který souvisí s tématem disertační práce, na publikování výsledků své tvůrčí práce a na vlastní zpracování disertační práce.
Do konce druhého roku studia skládá doktorand státní doktorskou zkoušku, kterou prokazuje široký rozhled a hluboké znalosti v oboru, souvisejícím s tématem disertační práce. K této zkoušce se musí přihlásit nejpozději do 30. dubna ve druhém roce svého studia. Státní doktorské zkoušce předchází zkouška z anglického jazyka.
Ve třetím a čtvrtém roce svého studia provádí doktorand potřebnou výzkumnou činnost, publikuje dosažené výsledky a zpracovává svoji disertační práci. Součástí studijních povinností v doktorském studijním programu je absolvování části studia na zahraniční instituci nebo účast na mezinárodním tvůrčím projektu s výsledky publikovanými nebo prezentovanými v zahraničí nebo jiná forma přímé účasti studenta na mezinárodní spolupráci, což je nutné doložit nejpozději při odevzdání disertační práce.
Doktorandi ve čtvrtém roce studia předkládají do konce zimního zkouškového období svému školiteli rozpracovanou disertační práci, který ji ohodnotí. Disertační práci doktorand odevzdává do konce 4. roku studia.
Student prezenční formy doktorského studia je v průběhu studia povinen absolvovat pedagogickou praxi, tj. působit v procesu výuky. Zapojení doktoranda do pedagogické činnosti je součástí jeho vědecké přípravy. Pedagogickou praxí doktorand získává zkušenosti v předávání poznatků a zdokonaluje prezentační dovednosti. Skladbu pedagogických aktivit (cvičení, laboratorní cvičení, vedení projektů apod.) určí doktorandovi vedoucí daného ústavu po dohodě se školitelem. Povinnost pedagogické praxe se nevztahuje na doktorandy-samoplátce a na doktorandy v kombinované formě studia. Zapojení do výuky v rámci pedagogické praxe potvrdí po jejím splnění školitel v IS VUT.

1. Approximate symbolic analysis of large systems

   Symbolic analysis allows to describe the behavior of an electronic circuit in the form of a symbolic expression. The project is focused on generating sufficiently simple and interpretable expressions that can be used by circuit designers. Due to the exponential growth of the expression complexity, the interpretability of exact expressions is lost even for circuits containing three transistors. An interesting way to get smaller expressions is to loosen accuracy to achieve simplicity, i.e. to use approximate symbolic analysis [1]. The method mimics designer's work, when they simplify the circuit model by omitting negligible phenomena (e.g. by omitting stray-capacitance currents on low frequencies, etc.). Despite the past development, there is still no widely accepted method of approximate symbolic analysis applicable to the circuits of practical size. Our previous results show a great potential of methods based on circuit model simplification (Simplification Before Generation class - SBG) [2]. During the first phase the research effort will be focused on improving the process of identification of negligible terms based on the Corner Analysis for checking the simplified expression validity on a certain interval of component parameters. The process should be modified to allow the simplification control based on inequalities following from typical device parameters or designer’s previous knowledge. The further modification of SBG will be aimed at suppressing mathematical nature of existing algorithms, which try to eliminate as many insignificant symbolic terms as possible, which results in problematic interpretability of obtained expressions [3]. The solution should be based on a graph-transformation method [2] and techniques of predictive modeling [4]. A potential candidate should have knowledge of electronic circuit theory, numerical mathematics and graph theory. Furthermore, knowledge of programming (C/C++, Python) is required, because it is expected that the developed methods will be implemented in the symbolic simulator SNAP. References [1] M. Fakhfakh, E. Tlelo-Cuautle, F. V. Fernández, Design of Analog Circuits through Symbolic Analysis, Bentham Science, 2012. [2] Z. Kolka, M. Vlk, M. Horák, “Topology Reduction for Approximate Symbolic Analysis,” Radioengineering, 2011, vol. 20, no. 1, pp. 252-256. [3] G. Shi, “Topological Approach to Symbolic Pole–Zero Extraction Incorporating Design Knowledge,” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. 36, no. 11, pp. 1765-1778, Nov. 2017. [4] D. Biolek, Z. Kolka, V. Biolková, Z. Biolek, S. Kvatinsky, “(V)TEAM for SPICE Simulation of Memristive Devices With Improved Numerical Performance,” IEEE Access, 2021, vol. 9, no. 2, pp. 30242-30255.

   Školitel: Kolka Zdeněk, prof. Dr. Ing.

2. Birdcage RF coil for MRI

   Magnetic resonance imaging (MRI) has gained increased interest as a diagnostic and research imaging technique. It offers very good soft tissue contrast at high spatial and temporal resolution [1]. MRI is able to move beyond anatomical imaging and it also allows visualizing metabolic functions and chemical processes. Basic components for MRI examination are MRI RF coils which are used for excitation or reception of the magnetic resonance signal. This project is focused on the research of a birdcage RF coil for MRI [2]-[3]. The main attention is concentrated on the improvement of magnetic field homogeneity in the transmission mode and signal to noise ratio in the receiving mode of a birdcage coil [2]-[4]. To manage this goal, extensive full-wave modeling and optimization in commercial software (e.g. CST Studio Suite or ANSYS HFSS) has to be carried considering relevant numerical phantoms. It is expected that a coil will be equipped by a self-adaptive technique allowing its adaptation for desired operating frequency band. As a result, a novel concept of a birdcage RF coil should be proposed. References [1] B. Gruber.; M. Froeling; T. Leiner; D. W. J. Klomp. RF coils: A practical guide for nonphysicists. Journal of Magnetic Resonance Imaging, 2018, vol. 48, no. 3, p. 590–604. [2] S. F. Ahmad; Y.C Kim; I. C. Choi; H. D. Kim. Recent Progress in Birdcage RF Coil Technology for MRI System. Diagnostics. 2020, vol. 10, no.12. [3] C.E. Hayes; W. A. Edelstein; J.F. Schenck; O. M. Mueller; M. Eash, An efficient, highly homogeneous radiofrequency coil for whole body NMR imaging at 1.5T. Journal of Magnetic Resonance. 1985, vol. 63, p. 622–628. [4] M. Vít; M. Burian; Z. Berková; J. Láčík; O. Sedláček; R. Hoogenboom; Z. Raida; D. Jirák, A broad tuneable birdcage coil for mouse 1H/19F MR applications. Journal of Magnetic Resonance, 2021, vol. 329, no. 1.

   Školitel: Láčík Jaroslav, doc. Ing., Ph.D.

3. Body-centric structure design by deep learning

   In the project, an artificial intelligence is going to be exploited for the genuine design of body-centric components to be used for sensing vital functions, in-body and on-body communication. Substituting a human intelligence by an artificial one in the design of body-centric systems should be the ambition of the research. A deep learning which mimics a human designer of such components has not been described in the open literature yet. Exploitation of deep learning for reverse engineering and the design of components operating in highly stochastic environments should be further contributions of the project. In preliminary research, we used deep structures for the classification of planar microwave structures. In consecutive steps, we identified edges on a photograph of a planar filter, estimated an equivalent circuit of a corresponding filter and revealed the type and the order of filtering structure. That way we proved that deep reverse models can be created. Deep learning is expected to be applied to the design of selected body-centric structures as endoscopic capsules or wireless epidermal devices. Numerical models of structures will be developed in commercial solvers, experiments will be implemented in cooperation with Faculty of Biomedical Engineering, CTU Prague. Deep learning algorithms are expected to be programmed in Python using available libraries (PyTorch, Keras, TensorFlow).

   Školitel: Raida Zbyněk, prof. Dr. Ing.

4. Body-centric structure design by deep learning

   In the project, an artificial intelligence is going to be exploited for the genuine design of body-centric components to be used for sensing vital functions, in-body and on-body communication. Substituting a human intelligence by an artificial one in the design of body-centric systems should be the ambition of the research. A deep learning which mimics a human designer of such components has not been described in the open literature yet. Exploitation of deep learning for reverse engineering and the design of components operating in highly stochastic environments should be further contributions of the project. In preliminary research, we used deep structures for the classification of planar microwave structures. In consecutive steps, we identified edges on a photograph of a planar filter, estimated an equivalent circuit of a corresponding filter and revealed the type and the order of filtering structure. That way we proved that deep reverse models can be created. Deep learning is expected to be applied to the design of selected body-centric structures as endoscopic capsules or wireless epidermal devices. Numerical models of structures will be developed in commercial solvers, experiments will be implemented in cooperation with Faculty of Biomedical Engineering, CTU Prague. Deep learning algorithms are expected to be programmed in Python using available libraries (PyTorch, Keras, TensorFlow).

   Školitel: Raida Zbyněk, prof. Dr. Ing.

5. Deep Learning Based Lossy Compression for 180°/360° Images

   Nowadays, popularity of 180°/360° images enabling to capture a panoramic (or omnidirectional) view of an environment has been rapidly increasing in many fields (e.g. industry and medicine) [1]. Such pictures with many characteristics result in high resolution leading to acquisition, transmission and storing based problems. Hence, coding algorithms with appropriate settings to compress the image content are necessary. Emerging still image compression algorithms, like JPEG XL, HEIF and AVIF, were not originally developed for compression of omnidirectional images. In recent years a comprehensive investigation into the possibility of using deep learning (DL) techniques for efficient image compression can be witnessed [2]. This work focuses on the development of DL-based lossy image compression technique for 180°/360° images. Attention should be devoted to study of factors having the highest influence on the quality of the compressed images in such format. Among others, it is expected that different objective and subjective approaches [1], [2] will be used for this purpose. The DL-based compression algorithm should find the tradeoff between complexity, accuracy and efficiency (e.g. time for the training) [2]. Following the development of DL-based compression technique, its compression performance should be assessed and compared to existing solutions. It is assumed that publicly available [3] and self-created dataset will be used for training the DL-based architectures. The DL algorithm (or algoritms) is expected to be programmed in Python using available libraries (PyTorch, Keras, TensorFlow) and should be publicly available for research community. [1] L. Polak, J. Kufa and T. Kratochvil, “On the Compression Performance of HEVC, VP9 and AV1 Encoders for Virtual Reality Videos,” in Proc. Of Int. Symp. Of BMSB 2020, Oct. 2020, pp. 1-5. DOI: 10.1109/BMSB49480.2020.9379878 [2] J. Ascenso et al. "Learning-based image coding: early solutions reviewing and subjective quality evaluation," Optics, Photonics and Digital Technologies for Imaging Applications VI. Vol. 11353. International Society for Optics and Photonics, Apr. 2020. DOI: 10.1117/12.2555368 [3] J. Gutiérrez et al., “Toolbox and dataset for the development of saliency and scanpath models for omnidirectional/360° still images,” Signal Processing: Image Communication, vol. 69, pp. 35-42, Nov. 2018. DOI: 10.1016/j.image.2018.05.003

   Školitel: Polák Ladislav, doc. Ing., Ph.D.

6. Deep Learning Based Prediction of Quality of Experience for Virtual Reality Videos

   Nowadays, interest about technology of virtual reality (VR) in multimedia systems is continuously increasing. In such systems, ensuring of excellent Quality of Experience (QoE) for VR videos (180 and 360-degree) will become a very important task in the future [1]. QoE can be influenced by many factors (e.g. compression algorithms, viewing conditions). Definition and prediction of these factors for VR videos become one of the key questions to research. Deep learning (DL) based technologies can be a suitable candidate to address this challenge [2]. This work focuses on the development of DL-based algorithm to predict QoE for VR videos. Attention should be devoted to study of factors having the highest influence on the QoE of VR videos. Among others, it is expected that different objective and subjective approaches [1] will be used for this purpose. The DL-based prediction algorithm should find the tradeoff between complexity, accuracy and efficiency (e.g. time for the training) [3]. Accuracy of the proposed DL-based prediction mechanism should be compared with conventional ones and evaluated in detail. The DL algorithm (or algorithms) is expected to be programmed in Python using available libraries (PyTorch, Keras, TensorFlow) and should be publicly available for research community. [1] L. Polak, J. Kufa and T. Kratochvil, “On the Compression Performance of HEVC, VP9 and AV1 Encoders for Virtual Reality Videos,” in Proc. Of Int. Symp. Of BMSB 2020, Oct. 2020, pp. 1-5. DOI: 10.1109/BMSB49480.2020.9379878 [2] Ch.-F. Hsu, T.-H. Hung and Ch.-H. Hsu, “Optimizing Immersive Video Coding Configurations Using Deep Learning: A Case Study on TMIV,” ACM Transactions on Multimedia Computing, Communications, and Applications, vol. 18, no. 1, pp. 1-25, Jan. 2022. DOI: 10.1145/3471191 [3] X. Feng, Y. Liu and S. Wei, "LiveDeep: Online Viewport Prediction for Live Virtual Reality Streaming Using Lifelong Deep Learning," in Proc. of Conf. on VR, March 2020, pp. 800-808. DOI: 10.1109/VR46266.2020.00104

   Školitel: Polák Ladislav, doc. Ing., Ph.D.

7. Electromagnetic cloaks

   Electromagnetic cloaks are structures that aim to reduce the reflectivity of the objects that surround or directly make them invisible. The theory of transformation electromagnetics and artificial materials are often exploited for their design [1], [2]. This project is focused on the research of electromagnetic cloaks of desired properties. The main attention should be concentrated on the development of methods for the design of artificial materials/surfaces with required electromagnetic properties. The outputs of the project should find application in the fields of antenna technology, security or defence applications [1]-[3]. After studying the current state of the art, the attention should be concentrated on the development of methods for the design of artificial materials/surfaces with the required electromagnetic properties based on machine learning approaches. For this part of the solution, the use of commercial software allowing full-wave modelling (e.g. CST Studio Suite, ANSYS HFSS) and the MATLAB program is assumed. Consequently, the newly created methods will be used for the design of selected reference structures whose properties will be verified experimentally. References [1] D. Kwon; D. H. Werner, Transformation Electromagnetics: An Overview of the Theory and Applications. IEEE Antennas and Propagation Magazine, 2010, vol. 52, no. 1, pp. 24-46. [2] D. H. Werner; D. Kwon, Transformation Electromagnetics and Metamaterials: Fundamental Principles and Applications. Springer, London, 2014. [3] G. Moreno et al., Wideband Elliptical Metasurface Cloaks in Printed Antenna Technology. IEEE Transactions on Antennas and Propagation, 2018, vol. 66, no. 7, pp. 3512-3525.

   Školitel: Láčík Jaroslav, doc. Ing., Ph.D.

8. Elektromagnetická analýza časově proměnlivých metapovrchů

   Časoprostorové elektromagnetické charakteristiky metapovrchů, jejichž materiálové vlastnosti jsou časově invariantní, byly detailně studovány a jsou dobře známé. Náročné požadavky kladené na 6G komunikační systémy však vyžadují nekonvenční inteligentní technologie, které by překonaly fyzikální limity standardních časově invariantních systémů. Perspektivní technologie schopné naplnit tyto požadavky jsou založené na adaptivních inteligentních metapovrších, jejichž elektromagnetické vlastnosti se mění v závislosti na proměnlivém okolním prostředí. Jelikož současné metody modelování elektromagnetického pole nedovolují jejich efektivní analýzu, výsledkem výzkumu budou fundamentálně nové výpočetní a analytické přístupy, které tak poskytnou klíčové nástroje pro návrh inteligentních časově proměnlivých metapovrchů. S ohledem na jejich bezkonkurenční výpočetní efektivitu, zvláštní pozornost bude věnována analytickým řešením a numerickým metodám pro řešení integrálních rovnic v časové oblasti, které jsou založeny na Cagniard-DeHoop metodě.

   Školitel: Štumpf Martin, doc. Ing., Ph.D.

9. Human activity monitoring from 5G/6G or new generation WiFi transmissions

   Monitoring of user activities using cameras or specialized radar-based equipment has found its use e.g. for monitoring of elderly people or to control the electronic entertainment equipment. The use of such approaches for large-scale monitoring in public spaces is problematic due to the privacy violation and need for specialized hardware. The aim of this PhD project is to investigate the influence of humans and their activities (e.g. pedestrians, bikers) on the statistical properties of millimeter-wave or sub-THz wireless channels in delay-Doppler domain and to search for suitable machine learning method able to classify the nature of user activities in scenarios of several users performing simultaneously a variety of distinct tasks. The investigated methods will be based on the processing of 5G New Radio, eventually 6G signals and should not require any specialized waveforms nor hardware. This PhD project includes both the experimental work with the mm-wave test-bed as well as data processing using state-of-the-art machine learning methods. There is an ongoing cooperation of the BUT group with several Austrian colleagues, e.g. Johannes Keppler University in Linz. The unique mm-wave test-bed in 60 GHz band available for experiments. [1] Ashleibta, A.M., Taha, A., Khan, M.A. et al. 5G-enabled contactless multi-user presence and activity detection for independent assisted living. Sci Rep 11, 17590 (2021). [2] Dubey, Anand & Santra, Avik & Fuchs, Jonas & Lübke, Maximilian & Weigel, Robert & Lurz, Fabian. (2021). A Bayesian Framework for Integrated Deep Metric Learning and Tracking of Vulnerable Road Users Using Automotive Radars. IEEE Access. 9. 10.1109/ACCESS.2021.3077690. [3] R. Marsalek, R. Zavorka, M. Pospisil, J. Vychodil, J. Gotthans and J. Blumenstein, "Human Activity Classification via Millimeter-Wave Channel Level Crossing Estimation," 2021 IEEE Microwave Theory and Techniques in Wireless Communications (MTTW), 2021, pp. 301-305

   Školitel: Maršálek Roman, prof. Ing., Ph.D.

10. Human activity monitoring from 5G/6G or new generation WiFi transmissions

    Monitoring of user activities using cameras or specialized radar-based equipment has found its use e.g. for monitoring of elderly people or to control the electronic entertainment equipment. The use of such approaches for large-scale monitoring in public spaces is problematic due to the privacy violation and need for specialized hardware. The aim of this PhD project is to investigate the influence of humans and their activities (e.g. pedestrians, bikers) on the statistical properties of millimeter-wave or sub-THz wireless channels in delay-Doppler domain and to search for suitable machine learning method able to classify the nature of user activities in scenarios of several users performing simultaneously a variety of distinct tasks. The investigated methods will be based on the processing of 5G New Radio, eventually 6G signals and should not require any specialized waveforms nor hardware. This PhD project includes both the experimental work with the mm-wave test-bed as well as data processing using state-of-the-art machine learning methods. There is an ongoing cooperation of the BUT group with several Austrian colleagues, e.g. Johannes Keppler University in Linz. The unique mm-wave test-bed in 60 GHz band available for experiments. [1] Ashleibta, A.M., Taha, A., Khan, M.A. et al. 5G-enabled contactless multi-user presence and activity detection for independent assisted living. Sci Rep 11, 17590 (2021). [2] Dubey, Anand & Santra, Avik & Fuchs, Jonas & Lübke, Maximilian & Weigel, Robert & Lurz, Fabian. (2021). A Bayesian Framework for Integrated Deep Metric Learning and Tracking of Vulnerable Road Users Using Automotive Radars. IEEE Access. 9. 10.1109/ACCESS.2021.3077690. [3] R. Marsalek, R. Zavorka, M. Pospisil, J. Vychodil, J. Gotthans and J. Blumenstein, "Human Activity Classification via Millimeter-Wave Channel Level Crossing Estimation," 2021 IEEE Microwave Theory and Techniques in Wireless Communications (MTTW), 2021, pp. 301-305

    Školitel: Maršálek Roman, prof. Ing., Ph.D.

11. Novel analog blocks, concepts and methods for sensing and processing of electrical and nonelectrical quantities

    The integrated circuits are very important for processing of signals from sensors and sensor readouts as a part of modern physical layer of communication systems [1], [2]. They offer significant minimization of system area and low power consumption. Therefore, these concepts are highly useful for biomedical applications (blood analysis – presence of various chemicals, bio-impedances measurement and evaluation, etc. [3], [4]), in mechanics (distance influences capacity) [5], etc. This topic includes study of utilization of discrete of-the-shelf as well as integrated active building cells and blocks (amplifiers, converters, generators, flip-flop circuits, etc.) and study of features of currently available types of sensors for various physical quantities. The recommendations, requirements, models, methodologies and specific solutions for various specific active sensor readouts and processing of signals are expected to be formulated for proposals of novel and advanced systems. The initial state of work concentrates on review of state of the art in discussed areas and results achieved at the workplace. It allows to find the most suitable specific topic (methodology, verification and measurement, modeling, discrete/integrated analog/mixed low-power or complex systems design) fitting to interests of candidate. These activities expect involvement in experimental work (in frame of projects of basic research – cooperation with research team including foreign experts) on design and implementation of integer-order as well as fractional-order circuits [4], modules (sensing readouts) [5] and components in discrete or integrated form and writing and dissemination of publications. This specialization offers significant enhancement of skills and competences in work with modern software tools (PSpice, Cadence Virtuoso/Spectre) of analog/mixed design approaches and further experience in detailed principles of advanced circuit solutions including cooperation on design of application specific integrated circuit. References [1] R. Sotner, J. Jerabek, L. Polak, J. Petrzela, W. Jaikla and S. Tuntrakool, “Illuminance Sensing in Agriculture Applications Based on Infra-Red Short-Range Compact Transmitter Using 0.35 um CMOS Active Device.” IEEE Access, vol. 8, pp. 18149-18161, 2020, doi: 10.1109/ACCESS.2020.2966752 [2] R. Sotner, L. Polak, J. Jerabek, “Low-cost remote distance and height sensing analog device for laboratory agriculture environments.” Measurement Science and Technology, online first, 2022, doi: 10.1088/1361-6501/ac543c [3] C. Vastarouchas, C.Psychalinos, A.S. Elwakil, A.A.Al-Ali, “Novel Two-Measurements-Only Cole-Cole Bio-Impedance Parameters Extraction Technique.” Measurement, vol. 131, pp. 394–399, 2019. doi: 10.1016/j.measurement.2018.09.008 [4] S. Kapoulea, C. Psychalinos, A. S. Elwakil, “Realization of Cole-Davidson function-based impedance models: Application on Plant Tissues.” Fractal and Fractional Journal, vol. 4, 54, 2020. doi: 10.3390/fractalfract4040054 [5] L. Polak, R. Sotner, J. Petrzela, J. Jerabek, “CMOS Current Feedback Operational Amplifier-Based Relaxation Generator for Capacity to Voltage Sensor Interface.” Sensors, vol. 18, 4488, 2018. doi: 10.3390/s18124488

    Školitel: Šotner Roman, doc. Ing., Ph.D.

12. Perspective methods for precise positioning of people and wireless devices in an indoor environment

    Nowadays, there are numerous methods to monitor, track and localize people and wireless devices in indoor environments. In the future, due to new emerging wireless communication systems (for instance the field of Internet-of-Things or Low-Power Wide Area Networks – LPWAN), it is assumed that current localization methods and techniques will need improvement or extension. From this point of view, utilization of machine learning and deep learning (ML and DL) techniques are among perspective solutions [1]-[3]. This dissertation thesis focuses on advanced methods and approaches for precise localization of people and wireless devices in an indoor environment. Development and realization of methods and approaches should be based on techniques evaluating parameters like RSSI, ToA and AoA [1]. Utilization of ML and DL-based approaches to improve efficiency and accuracy of localization methods in an indoor environment is assumed [3]. Testing and verification of the proposed methods and approaches by a set of measurements under laboratory and real conditions is an inseparable part of this work. It is assumed that publicly available and self-created dataset will be used for training the DL-based architectures [3]. The DL algorithm (or algorithms) is expected to be programmed in Python using available libraries (PyTorch, Keras, TensorFlow) and should be publicly available for research community. The final DL algorithm must find tradeoff between complexity, accuracy and efficiency. [1] F. Zafari, A. Gkelias and K. K. Leung, "A Survey of Indoor Localization Systems and Technologies," IEEE Communications Surveys & Tutorials, vol. 21, no. 3, pp. 2568-2599, Thirdquarter 2019, DOI: 10.1109/COMST.2019.2911558 [2] J. Pelant et al., "BLE device indoor localization based on RSS fingerprinting mapped by propagation modes," 2017 27th International Conference Radioelektronika (RADIOELEKTRONIKA), 2017, pp. 1-5, DOI: 10.1109/RADIOELEK.2017.7937584 [3] L. Polak et al. Received Signal Strength Fingerprinting-Based Indoor Location Estimation Employing Machine Learning. Sensors, 2021, vol. 21, no. 13, pp. 1–25. DOI: 10.3390/s21134605

    Školitel: Polák Ladislav, doc. Ing., Ph.D.

13. Pokročilé algoritmy pro trénování neuronových sítí

    Disertační práce se zaměřuje na vývoj fundamentálně nových technik optimalizace pro trénování neuronových sítí. Detailní popis viz anglická verze.

    Školitel: Kadlec Petr, doc. Ing., Ph.D.

14. Pokročilé optimalizační algoritmy pro clustering dat

    Disertační práce se zaměřuje na vývoj fundamentálně nových technik optimalizace pro řešení problému clusteringu. Detailní popis viz anglická verze.

    Školitel: Kadlec Petr, doc. Ing., Ph.D.

15. Tunability range extension in electronically adjustable circuits

    Decreasing value of power supply voltage creates limited conditions for electronic tuning of circuits (for example active filters and oscillators) in comparison to standard current systems operating with high DC power supply. The main task of this work focuses on research and study of methods of electronic control of applications (for example filters and oscillators) working as components of modern communication systems. Suitable combination of features for control of active elements [1] and change of character of dependence (e.g. oscillation frequency vs. adjustable parameter) [2], as well as popular fractional-order approaches [3] serve for substantial improvement of tunability range and features of circuits for application in communication systems, in various methods of measurement, etc. The initial stage of work concentrates on review of state of the art of methods and approaches for tunability enhancement achieved at the workplace as well as on principles of advanced active elements suitable for these purposes. Further activities target on formulation of novel approaches and experimentally verified methodologies using discrete/integrated analog circuits (modern software tools as PSpice, Cadence Virtuoso/Spectre) including cooperation on design of application specific integrated circuit (experience in detailed principles of advanced circuit solutions). Cooperation with international research group (in frame of projects of basic research) on writing and dissemination of publications is possible and expected. References [1] R. Sotner, J. Jerabek, N. Herencsar and J. Petrzela, "Methods for Extended Tunability in Quadrature Oscillators Based on Enhanced Electronic Control of Time Constants." IEEE Transactions on Instrumentation and Measurement, vol. 67, no. 6, pp. 1495-1505, 2018. doi: 10.1109/TIM.2018.2799058. [2] R. Sotner, J. Jerabek, J. Petrzela, O. Domansky, W. Jaikla and T. Dostal, "Exponentially tunable voltage controlled quadrature oscilator." 40th International Conference on Telecommunications and Signal Processing (TSP), 2017, pp. 302-306, doi: 10.1109/TSP.2017.8075992. [3] R. Sotner, J. Jerabek, L. Polak, L. Langhammer, H. Stolarova, J. Petrzela, D. Andriukaitis, A. Valinevicius, “On the performance of electronically tunable fractional-order oscillator using grounded resonator concept.” AEU - International Journal of Electronics and Communications, vol. 129, pp. 1-17, 2021. doi: 10.1016/j.aeue.2020.153540

    Školitel: Šotner Roman, doc. Ing., Ph.D.

16. Unconventional approaches for transmission lines modelling and simulation

    In today’s high-speed mixed electronic systems the transmission lines play important role as interconnects inside and/or among their subsystems due to wave effects occuring at frequencies above hundreds of megahertz [1]. Real interconnects have to moreover be considered as lossy, often nonuniform, and in some special cases also nonlinear. The losses of transmission lines, especially a skin effect and a dielectric loss, are frequency dependent which is becoming more significant for high frequencies. The transmission lines are commonly described by systems of telegraph equations. As is known, however, a number of physical effects cannot be described with sufficient accuracy by using classical integer-order differential equations. In this field, the theory of fractional-order differential equations (FDE) and corresponding models can be utilized with advantage for their more credible description [2]-[4]. Besides, as a result of imperfect manufacturing processes the real interconnects suffer from random disturbances in values of their parameters which should be taken into account in devices designs. There are various approaches to characterize these random variations, e.g. Monte Carlo or polynomial chaos methods [5]. One of the promissing direction is to apply the theory of stochastic differential equations (SDE) [6]. So far most models and techniques were based on the application of ordinary SDEs [7], [8], an open possibility is to utilize partial SDEs as well. The topic is a part of broader signal integrity issues being solved in high-speed electronic systems [1]. Currently the above approaches and corersponding models are usually considered separatelly. The aim of the thesis is to develop models and methods for computer simulation of transmission lines which would be able to take into account both fractional nature and stochastic variability of their parameters. It is supposed that Matlab and PSpice programs will widely be utilized at the solution, therefore, their knowledge is required at candidates, as well as their interest of mathematical modelling. [1] S. H. Hall, H. L. Heck, Advanced Signal Integrity for High-Speed Digital Designs, New York: John Wiley & Sons, 2011. [2] Fractional-Order Modeling of Dynamic Systems with Applications in Optimization, Signal Processing, and Control, Edited by Ahmed G. Radwan, Farooq Ahmad Khanday and Lobna A. Said, Academic Press, 2021. [3] N. Al-Zubaidi R-Smith, A. Kartci, L. Brancik, “Application of Numerical Inverse Laplace Transform Methods for Simulation of Distributed Systems with Fractional-Order Elements,” Journal of Circuits, Systems, and Computers JCSC, vol. 27, no. 11, p 1-25, 2018. [4] S. M. Cvetianin, D. Zorica, and M. R. Rapaić, “Generalized time-fractional telegraphers equation in transmission line modeling,” Nonlinear Dynamics, vol. 88, no. 2, p. 1453-1472, 2017. [5] P. Manfredi, D. V. Ginste, I. S. Stievano, D. De Zutter and F. G. Canavero, "Stochastic transmission line analysis via polynomial chaos methods: an overview," IEEE Electromagnetic Compatibility Magazine, vol. 6, no. 3, p. 77-84, 2017. [6] S. Särkkä, A. F. Solin, Applied Stochastic Differential Equations. Cambridge: Cambridge University Press, 2019. [7] A. Zjajo, Q. Tang, M. Berkelaar, J. P. de Gyvez, A. Di Bucchianico, N. van der Meijs, ”Stochastic analysis of deep-submicrometer CMOS process for reliable circuits designs,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 58, no. 1, 2011, p. 164-175. [8] L. Brancik, E. Kolarova, “Simulation of multiconductor transmission lines with random parameters via stochastic differential equations approach,” SIMULATION-Transactions of the Society for Modeling and Simulation International, vol. 92, no. 6, p. 521-533, 2016.

    Školitel: Brančík Lubomír, prof. Ing., CSc.

2. kolo (podání přihlášek od 01.07.2022 do 31.07.2022)

1. Methods for testing of COTS semiconductor components radiation hardness with respect to space missions

   In recent years more and more commercial companies and research institutions are willing to build new satellites and space probes. In case of their electronic equipment, so called "space grade" parts are traditionally used. Such components are manufactured with respect to harsh environment (space), where they are intended to work. Among the most demanding factors are temperature cycling, vacuum, mechanical stress and ionizing radiation. Regarding the radiation, space grade components are guaranteed to withstand certain intensities and doses of ionizing radiation before they cease functioning. However, such rugged components are very expensive and often quite obsolete in terms of overall performance when compared to commercial off-the-shelf (COTS) components. Thus many companies are trying to utilize COTS components in their space probe and satellite design to improve electronic system performance (increase computing power, decrease power consumption). FPGAs are especially interesting from this point of view. To ensure sufficient reliability of the whole electronic system, all the COTS components used in the design shall be tested on radiation hardness. Traditionally, this is achieved using a Cobalt-60 gamma ray source, as it is widely available and easy to use. However, the space environment is more complex, energy spectrum of the Cobalt-60 is not a perfect match. The aim of the research is to search for alternative methods of electronic components testing, for example utilizing widespread proton accelerators and their parasitic radiative field. It is expected that the methodology will be verified on an FPGA platform. To support the project, we have active cooperation with Masaryk University Brno, VF inc., Nuclear Research Centre Rez, and Department of Nuclear Reactors at CVUT Prague. [1] VELAZCO, R., MCMORROW, D, ESTELA , J. (Editors). Radiation Effects on Integrated Circuits and Systems for Space Applications. Springer Nature Switzerland AG 2019. ISBN 978-3-030-04660-6. [2] Y. Kimoto, N. Nemoto, H. Matsumoto, et al., Space radiation environment and its effects on satellites: analysis of the first data from TEDA on board ADEOS-II. IEEE Trans. Nucl. Sci.52(5),1574–1578 (2005) [3] EIA/JESD57, Test Procedures for the Manegement of Single-Event Effects in Semiconductor Devices from Heavy-Ion Irradiation (EIA/JEDEC Standard, Nov. 2017, available at: https://www.jedec.org/standards-documents/docs/jesd-57) [4] ASTM F 1192-11, Standard Guide for the Measurement of Single Event Phenomena (SEP) Induced by Heavy Ion Irradiation of Semiconductor Devices (ASTM Standard, West Conshohocken, PA, 2006) [5] MIL-STD-750-1, Environmental Test Methods for Semiconductor Devices (Department of Defense Test Method Standard, USA, 2012)

   Školitel: Kolka Zdeněk, prof. Dr. Ing.

1. kolo (podání přihlášek od 01.04.2022 do 15.05.2022)

1. Deep Learning for Classification of Coexistence of Wireless Communication Systems

   In the future, different wireless communication systems can share common radiofrequency (RF) bands. Such a so called coexistence of these systems can be critical (a partial or full loss of wireless services, provided by communication systems) or non-critical (communication systems can coexist without significant performance degradation) [1]-[3]. Hence, predicting and coordinating the coexistence of these systems will be an important task. Deep learning (DL) based technologies can be a suitable candidate to address such challenges [4]. This work focuses on the development of DL-based algorithm for classification of coexistence scenarios between different wireless communication systems in terms of RF signals. Attention should be devoted to the study of parameters having the highest influence on the character of the interfering signal (e.g. idle signal, modulation scheme, type of modulation). Many parameters enable the DL-based architectures to learn more features from the data [5]. Hence, the DL algorithm must find tradeoff between complexity, accuracy and efficiency. It is assumed that publicly available and self-created dataset will be used for training the DL-based architectures. The DL algorithm (or algorithms) is expected to be programmed in Python using available libraries (PyTorch, Keras, TensorFlow) and should be publicly available for research community. [1] Y. Han, E. Ekici, H. Kremo and O. Altintas, “Spectrum sharing methods for the coexistence of multiple RF systems: A survey,” Ad Hoc Networks, vol. 53, pp. 53-78, Dec. 2016. DOI: 10.1016/j.adhoc.2016.09.009 [2] A. M. Voicu, L. Simić and M. Petrova, "Survey of Spectrum Sharing for Inter-Technology Coexistence," IEEE Communications Surveys & Tutorials, vol. 21, no. 2, pp. 1112-1144, Secondquarter 2019, DOI: 10.1109/COMST.2018.2882308 [3] L. Polak and J. Milos, “Performance analysis of LoRa in the 2.4 GHz ISM band: coexistence issues with Wi-Fi,” Telecommunication Systems, vol. 74, no. 3, pp. 299-309, July 2020. DOI: 10.1007/s11235-020-00658-w [4] Y. Shi, K. Davaslioglu, Y. E. Sagduyu, W. C. Headley, M. Fowler and G. Green, "Deep Learning for RF Signal Classification in Unknown and Dynamic Spectrum Environments," In Proc of. Int. Symp. DySPAN, Nov. 2019, pp. 1-10, DOI: 10.1109/DySPAN.2019.8935684 [5] K. Pijackova and T. Gotthans, "Radio Modulation Classification Using Deep Learning Architectures," In Proc of. 31st Int. Conf. Radioelektronika, Apr. 2021, pp. 1-5, DOI: 10.1109/RADIOELEKTRONIKA52220.2021.9420195

   Školitel: Polák Ladislav, doc. Ing., Ph.D.

2. Millimeter wave channel characterization using machine learning

   Steadily growing number of communication devices per area and increasing quality of services require allocation of more frequency resources. Millimeter wave (MMW) frequencies between 30 and 300 GHz have been attracting growing attention as a possible candidate for next-generation broadband cellular networks. Specific limitations of MMW signal propagation, extremely large bandwidth and time variable environment caused by mobile users connected to a backhaul networks traveling in rugged municipal environments create unprecedented challenges to the development of broadband communication systems using advanced technologies for eliminating the undesirable time varying channel features. The general objective of the project is measurement and modelling of the broadband time varying MMW channels between mobile users and infrastructure in delay and spatial domain and extension of our previous research focused on the characterization of intra-vehicle and V2X channels for the purposes of stochastic channel modeling [1]. The parametrization of channel models needs an accurate extraction of the channel parameters such as number, position and amplitudes of multipath components (MPC), clusters, LOS, and NLOS components, etc. in the delay and spatial domains from measured data. Real time capturing of MPCs in a very wide spatial angle is provided, for example, by measuring systems with a fast-spinning antenna. However, such a system produces a huge amount of data. Thus, to get all the MPC related parameters, some automated algorithm is needed. Such algorithms are based for example on identifying the changes in the slope of a channel impulse response or generally on parameter threshold-based identification. Due to the limited accuracy and reliability of many of these methods, we are going to use machine learning (ML) techniques such as Gaussian Mixture Model or K-means algorithm for gathering MPCs with similar parameters behavior [2]. Further the project also envisages the use of supervised ML such as Deep Neural Networks or Support Vector Machine to predict and estimate the channel parameters and examine large and small-scale fading including parameters such as path loss, delay path loss exponent, Doppler spread, angle of arrival, and other variables describing the channel [3]. The above algorithms are expected to be implemented using Machine Learning Workflow with Keras, Tensorflow, and Python [4]. An alternative implementation in MATLAB also possible. The student will be a member of the international team of scientists from Brno University of Technology, TU Vienna, Austrian Institute of Technology Vienna, University of Southern California, National Institute of Technology Durgapur India, and Military University of Technology Warsaw. [1] E. Zöchmann, M. Hofer, M. Lerch, S. Pratschner, L. Bernado, J. Blumenstein, S. Caban, S. Sangodoyin, H. Groll, T. Zemen, A. Prokeš, M. Rupp, A. Molisch, C. Mecklenbräuker, Position-Specific Statistics of 60 GHz Vehicular Channels During Overtaking. IEEE Access, 2019, vol. 7, no. 1, p. 14216-14232. [2] S. M. Aldossari, K.C. Chen, Machine Learning for Wireless Communication Channel Modeling: An Overview, Wireless Personal Communications, 2019, 106, p. 41 – 70. [3] R. A. Osman, S. N. Saleh, Y. N. M. Saleh, M. N. Elagamy, Enhancing the Reliability of Communication between Vehicle and Everything (V2X) Based on Deep Learning for Providing Efficient Road Traffic Information. Applied Science, 2021, vol. 11, art. no. 11382. [4] C. A. Mattmann, Machine Learning with TensorFlow, Second Edition, Manning Publications, 2021.

   Školitel: Prokeš Aleš, prof. Ing., Ph.D.

3. Multi-objective synthesis of slot-based antennas

   At millimeter-wave frequencies, antennas are usually fed by substrate-integrated structures to create a compact geometry and minimize a parasitic radiation of feeders. When integrating, feeders in the substrate can be completed by additional components like filters, phase shifters, power dividers, switches etc. By etching slots to walls of substrate-integrated structures, slot antennas can be created. The designer of millimeter-wave antennas (antenna arrays) considers all objectives of the design comprising technical parameters, price and limitations defined by a user. In order to meet requirements, known partial solutions are combined and optimized to get as good match with objectives as possible. Within the project, the described general synthesis of slot antenna systems fed by substrate-integrated structures is going to be implemented in an algorithmic way: 1. Starting from the requested parameters of the antenna to be designed, the optimum antenna configuration should be determined. Components of the designed structure are selected from a library minimizing dimensions and complexity. The structural optimization is performed. 2. Comprising conflicting requirements on antenna properties, the Pareto front of optimal solutions is computed by a multi-objective optimization algorithm; evolutionary approaches or swarm-intelligence ones are expected. In order to increase efficiency of the synthesis, tuning space-mapping approaches might be applied. The developed design procedure is going to be tested on practical examples from the area of vehicular applications.

   Školitel: Raida Zbyněk, prof. Dr. Ing.

4. Multi-objective synthesis of slot-based antennas

   At millimeter-wave frequencies, antennas are usually fed by substrate-integrated structures to create a compact geometry and minimize a parasitic radiation of feeders. When integrating, feeders in the substrate can be completed by additional components like filters, phase shifters, power dividers, switches etc. By etching slots to walls of substrate-integrated structures, slot antennas can be created. The designer of millimeter-wave antennas (antenna arrays) considers all objectives of the design comprising technical parameters, price and limitations defined by a user. In order to meet requirements, known partial solutions are combined and optimized to get as good match with objectives as possible. Within the project, the described general synthesis of slot antenna systems fed by substrate-integrated structures is going to be implemented in an algorithmic way: 1. Starting from the requested parameters of the antenna to be designed, the optimum antenna configuration should be determined. Components of the designed structure are selected from a library minimizing dimensions and complexity. The structural optimization is performed. 2. Comprising conflicting requirements on antenna properties, the Pareto front of optimal solutions is computed by a multi-objective optimization algorithm; evolutionary approaches or swarm-intelligence ones are expected. In order to increase efficiency of the synthesis, tuning space-mapping approaches might be applied. The developed design procedure is going to be tested on practical examples from the area of vehicular applications.

   Školitel: Raida Zbyněk, prof. Dr. Ing.

5. Neutron and gamma radiation spectroscopy using proportional detectors

   Currently Bonner spheres are used to measure field of low-energy neutrons (< 100 keV). This method is cumbersome and time-consuming, which limits its application. Utilization of proportional detectors is an alternative and very promising method suitable for measuring mixed fields of photons (gamma rays) and neutrons in energetic range from about 20 keV to 1 MeV. The main benefit of using the proportional detector is that it is capable of both energy measurement and particle discrimination (gamma / neutron) in wide energy range and in reasonably short time. However, so far there is only limited success in the processing of proportional detector output signals, namely discrimination of gamma/neutron particles and energy estimation of those particles. The reason is that the response of the detector is dependent not only on energy and type of the particle, but also on the geometry of the detector and actual trajectory of the particle travelling through the detector. The aim of the thesis is to establish methods (algorithms) for particle discrimination and their respective energy measurement. Currently we have an experimental setup based on FPGA-based data acquisition board, which is intended for experimental data gathering. The data shall be processed in a PC to determine optimum method of energy measurement and particle discrimination. Machine learning techniques are one promising method that shall be (at least) considered. Later on, the FPGA should include those methods (algorithms) so that it can provide measurement results (energy / particle distribution) directly, acting as a stand-alone measurement device. To support the project, we have active cooperation with Masaryk University Brno, VF inc., Nuclear Research Centre Rez, and Department of Nuclear Reactors at CVUT Prague. Those institutions will provide equipment required for experiments with proportional detectors (gamma and neutron radiation sources). The measurement hardware (FPGA-based digitizer with preamplifier) is currently under development in frame of master thesis of student Ondrej Kolar. Ondrej is going to follow up with this topic on his PhD thesis. [1] KNOLL, Glenn F. Radiation Detection and Measurement. 3rd edition. Michigan: John Wiley & Sons, 2000. ISBN 0-471-07338-5. [2] LANGFORD, T.J., C.D. BASS, E.J. BEISE, H. BREUER, D.K. ERWIN, C.R. HEIMBACH a J.S. NICO. Event identification in 3He proportional counters using risetime discrimination. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment [online]. 2013, 717, 51-57 [cit. 2021-12-29]. ISSN 01689002. doi:10.1016/j.nima.2013.03.062 [3] HEEGER, K.M., S.R. ELLIOTT, R.G.H. ROBERTSON, M.W.E. SMITH, T.D. STEIGER a J.F. WILKERSON. High-voltage microdischarge in ultra-low background 3He proportional counters. IEEE Transactions on Nuclear Science [online]. 2000, 47(6), 1829-1833 [cit. 2021-12-29]. ISSN 0018-9499. doi:10.1109/23.914454

   Školitel: Kolka Zdeněk, prof. Dr. Ing.

6. Odhad časoprostorové funkce v mobilních scénářích

   Odhad prostorové funkce je úkol zpracování signálu, který nachází uplatnění mimo jiné v oblasti telekomunikací jako způsob odhadu mapy síly přijímaného signálu (RSS) a dalších prostorově se měnících metrik výkonu na základě pozorování senzorů. V této aplikaci umožňuje vysoce kvalitní odhad prostorové funkce efektivní alokaci síťových zdrojů. Cílem tohoto projektu je prozkoumat a formulovat metody pro odhad prostorových funkcí s ohledem na různé problémové aspekty. Tyto aspekty zahrnují mobilitu senzorů, nejistotu polohy senzoru, distribuovaný provoz senzoru a rozšířené informace o umístění senzoru. Kromě toho je třeba vzít v úvahu také časový rozměr scénáře odhadu. Důraz bude kladen na využití Gaussovských regresních procesů jako vysoce výkonné a všestranné metody pro odhad prostorových funkcí.

   Školitel: Poměnková Jitka, doc. RNDr., Ph.D.

7. Perspektivní bezdrátové optické komunikační systémy pro komunikační standard B5G a 6G

   Neustálý vývoj nových standardů pro bezdrátové komunikace klade stále větší nároky na jednotlivé komunikační komponenty z hlediska přenosové rychlosti, přenosové kapacity, bezpečnosti, univerzálnosti, škálovatelnosti a energetické efektivity. Uspokojení těchto požadavků vyžaduje pokročilý hardware a technologie pracující v nových spektrálních pásmech [1]. Optické bezdrátové komunikace, díky svým specifickým vlastnostem a neustálému vývoji, zastávají významné a nezastupitelné místo v moderních komunikačních technologiích ať už na krátké, nebo na dlouhé vzdálenosti. Plánovaná implementace optických bezdrátových spojů (FSO) v B5G a 6G standardech pro mobilní komunikace je impulsem pro hustší nasazení těchto spojů ve městech a v zastavěných oblastech. Rovněž se předpokládá nasazení pro dočasné (ad-hoc) sítě mezi UAV (Unnamed Aerial Vehicle) a pozemní stanicí. Cílem vědeckého projektu je výzkum v oblasti metod generování optického signálu a metod detekce v optických bezdrátových komunikačních systémech (FSO a VLC), které jsou implementovány ve standardu B5G, nebo jsou plánované ve standardu 6G. Výzkum bude zaměřen na zpracování signálu v optických transceiverech. Budou analyzovány nové pokročilé typy modulací a kanálového kódování. Experimentální práce bude zaměřena na porovnání vybraných typu modulátorů a detektorů. Cílem výzkumu je potlačení negativního vlivu atmosféry [2]-[4] na přenos optického signálu, optimalizace přenosové technologie a zvýšení její spolehlivosti a dostupnosti. [1] PÄRSSINEN, A.; ALOUINI, M.; BERG, M.; KUERNER, T.; KYÖSTI, P.; LEINONEN, M. E.; MATINMIKKO-BLUE; M., MCCUNE, E.; PFEIFFER, U., and WAMBACQ, P. (Eds.). (2020). White Paper on RF Enabling 6G – Opportunities and Challenges from Technology to Spectrum [White paper]. (6G Research Visions, No. 13). University of Oulu. [2] BARCIK, P.; WILFERT, O.; DOBESCH, A.; KOLKA, Z.; HUDCOVA, L.; NOVAK, M.; LEITGEB, E. Experimental measurement of the atmospheric turbulence effects and their influence on performance of fully photonic wireless communication receiver. Physical Communication, 2018, vol. 31, no. 1, p. 212-217. ISSN: 1874-4907. [3] MAREK NOVAK; PETER BARCIK; PETR SKRYJA; ZDENEK KOLKA. Service Data Transmission System for Free Space Optics. In 20th Conference on Microwave Techniques COMITE 2021. Brno: IEEE, 2021. s. 1-4. ISBN: 978-1-6654-1454-8. [4] POLIAK, J.; BARCIK, P.; WILFERT, O. Diffraction Effects and Optical Beam Shaping in FSO Terminals. In Optical Wireless Communications - An Emerging Technology. Springer International Publishing Switzerland: Springer International Publishing, 2016. p. 1-21. ISBN: 978-3-319-30200- 3.

   Školitel: Hudcová Lucie, doc. Ing., Ph.D.

8. Spatio-temporal data analysis by neural networks

   The thesis project focuses on spatio-temporal networks and their use in visual pattern recognition. Spatio-temporal data contains temporal and spatial information simultaneously, and therefore such dependencies are often used together in prediction models. The models are used on a variety of forecasting problems from different areas, such as traffic flow prediction, weather and climate environment, social media, video segmentation, people migration, etc. In the project we will rely on deep learning to deal with spatio-temporal data. Various network structures were extended from the Recurrent Neural Network to train temporal patterns from historical sequence information to predicting spatio-temporal data [1]. The time-varying data are also represented by graphs or by Graph Convolutional Networks which alternate temporal and spatial convolutions [2]. The aim of this work is to study new representations for spatio-temporal graphs and to design new neural network architectures for data represented by this type of graphs. These models will be programmed in Python and compared with the state-of-the-art datasets for different applications. [1] Son, H.; Kim, S.; Yeon, H.; Kim, Y.; Jang, Y.; Kim, S.-E. Visual Analysis of Spatiotemporal Data Predictions with Deep Learning Models. Appl. Sci. 2021, 11, 5853. https://doi.org/10.3390/app11135853 [2] Yu, Bing & Yin, Haoteng & Zhu, Zhanxing. (2018). Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting. 3634-3640. 10.24963/ijcai.2018/505. [3] Mathe, Johan, et al. "PVNet: A LRCN architecture for spatio-temporal photovoltaic PowerForecasting from numerical weather prediction." arXiv preprint arXiv:1902.01453 (2019).

   Školitel: Frýza Tomáš, doc. Ing., Ph.D.

9. Syntéza neceločíselných přenosových struktur

   Disertační práce se zaměřuje na návrh, syntézu a analýzu přenosových struktur neceločíselného řádu (kmitočtové filtry a základní stavební bloky (integrátor, derivátor) neceločíselných řádů). Součástí návrhu je i schopnost elektronické změny dílčích parametrů navržených zapojení. Jako aktivní prvky jsou využívány zejména moderní obvody jako například proudové sledovače a zesilovače, transkonduktanční zesilovače, různé typy konvejorů a dále pak i složené aktivní prvky (VDTA, VDCC, CCTA atd.). Předmětem studie je i možnost elektronické rekonfigurace typu přenosové odezvy. Pro ověření správného návrhu a funkčnosti navržených zapojení budou použity simulace (na diskrétní integrované úrovni) a experimentálním měření s využitím dostupných aktivních prvků.

   Školitel: Langhammer Lukáš, Ing., Ph.D.

10. Turbulentní model vodního přenosového prostředí pro optické bezdrátové komunikace

    Optické komunikace pracující ve vodním prostředí (Underwater optical communications UWOC) jsou jedním z aktuálních a perspektivních směrů v bezdrátových komunikacích. Zásadní výhodou tohoto typu komunikace je přenos informace v reálném čase a vysoké přenosové rychlosti pro krátké vzdálenosti [1], [2]. Dosah UWOC linek je však výrazně omezen spektrálně závislými útlumovými vlastnostmi vodního prostředí. Zásadní vliv na přenos optického signálu má také proudění vodní masy (vodní turbulence), která závisí na mnoha parametrech prostředí (např. teplota vody, salinita, rychlost proudění vody, index lomu vody, hloubka vody, reliéf dna) [3], [4]. Cílem projektu je detailní analýza vodního prostředí s ohledem na proudění vodní masy a stanovení míry turbulence podvodního přenosového prostředí. Hlavním výstupem projektu bude turbulentní model vodního prostředí, který bude definovat změny v šíření optických svazků. Při analýze šíření optických svazků turbulentním prostředím je potřeba brát v úvahu například intenzitní profil svazku, jeho koherenci, pološířku svazku nebo vlnovou délku. Je potřeba také stanovit míru disperze optického signálu pro různé směry šíření optických svazků. Důležitým cílem projektu je stanovení limitů vodního prostředí pro dosah a přenosové rychlosti optických linek. [1] H. Kaushal and G. Kaddoum, "Underwater Optical Wireless Communication," in IEEE Access, vol. 4, pp. 1518-1547, 2016, doi: 10.1109/ACCESS.2016.2552538. [2] Z. Zeng, S. Fu, H. Zhang, Y. Dong and J. Cheng, "A Survey of Underwater Optical Wireless Communications," in IEEE Communications Surveys & Tutorials, vol. 19, no. 1, pp. 204-238, Firstquarter 2017, doi: 10.1109/COMST.2016.2618841. [3] C. T. Geldard, J. Thompson and W. O. Popoola, "Empirical Study of the Underwater Turbulence Effect on Non-Coherent Light," in IEEE Photonics Technology Letters, vol. 32, no. 20, pp. 1307-1310, 15 Oct.15, 2020, doi: 10.1109/LPT.2020.3020368. [4] S. Zhang, L. Zhang, Z. Wang, J. Quan, J. Cheng and Y. Dong, "On Performance of Underwater Wireless Optical Communications Under Turbulence," 2020 IEEE 17th Annual Consumer Communications & Networking Conference (CCNC), 2020, pp. 1-2, doi: 10.1109/CCNC46108.2020.9045458.

    Školitel: Hudcová Lucie, doc. Ing., Ph.D.

11. Vytěžování specifických informací z řečových signálů

    Řeč je nejpřirozenějším způsobem komunikace mezi lidmi [1]. Kromě přenosu zprávy obsahuje řeč (a zaznamenaný řečový signál) také znaky identifikující řečníka. Pro výzkum jsou zajímavé specifické rysy v řeči odrážející aktuální stav řečníka jako je nálada, stres, nedostatek spánku, alkohol atd., ale i informace o přítomnosti některých onemocnění, např. Parkinsonovy choroby. Zatímco sdělená zpráva a identita řečníka jsou obvykle dobře rozpoznatelné pouhým poslechem, hodně specifických informací zůstává skrytých. Lze je však odhalit pomocí počítačové analýzy řečových signálů. Výzkumné práce se budou řídit současným stavem techniky. V počáteční fázi bude potřeba zpracovat přehledovou studii o známých metodách získávání specifických informací z řečových signálů k rozpoznání negativních emocí [2], [3] a stavu vyhoření. Jádrem výzkumu bude odvození hlasivkových pulzů z řečového signálu a jejich zpracování v časové, frekvenční i kepstrální oblasti [4]. Bude vyvíjen optimální algoritmus pro získání efektivních hlasivkových pulzů. Úspěšnost a přesnost celkového přístupu k odhalování vybraných psychologických jevů bude testována na reálných řečových datech. [1] Amon, I. (2020) Die Macht der Stimme. Redline Verlag, München. [2] Chourasia, M., Haral, S., Bhatkar, S., & Kulkarni, S. (2021). Emotion recognition from speech signal using deep learning. In Intelligent Data Communication Technologies and Internet of Things, Springer, Singapore, pp. 471-481. [3] Fahad, M. S., Ranjan, A., Yadav, J., & Deepak, A. (2021). A survey of speech emotion recognition in natural environment. In Digital Signal Processing, 110, 102951. [4] Rabiner, L., & Schafer, R. (2011) Theory and Applications of Digital Speech Processing, Prentice Hall, London.

    Školitel: Sigmund Milan, prof. Ing., CSc.