<p>The Kerngen Module in MATLAB can be used to optimize a filter with regards to an ideal filter; while taking into consideration the weighting function and the spatial mask. To be able to remotely do these optimizations from a standard web browser over a TCP/IP network connection would be of interest. This master’s thesis covers the project of doing such a system; along with an attempt to graphically display three-dimensional filters and also save the optimized filter in XML format. It includes defining an appropriate DTD for the representation of the filter. The result is a working system, with a server and client written in the programming language PIKE.</p>

The Kerngen Module in MATLAB can be used to optimize a filter with regards to an ideal filter; while taking into consideration the weighting function and the spatial mask. To be able to remotely do these optimizations from a standard web browser over a TCP/IP network connection would be of interest. This master’s thesis covers the project of doing such a system; along with an attempt to graphically display three-dimensional filters and also save the optimized filter in XML format. It includes defining an appropriate DTD for the representation of the filter. The result is a working system, with a server and client written in the programming language PIKE.

The filter complexity in the multi-stage decimation system of a Δ-Σ ADC increases progressively as one moves to higher stages of decimation due to the fact that the input word length of the higher stages also increases progressively. The main motivation for this thesis comes from the idea of investigating a way, to reduce the input word length in the later filter stages of the decimation system which could reduce the filter complexity. To achieve this, we use a noise-shaping loop between the first and later stages so that the input word length for the later stages remains smaller than in the case where we do not use the noise-shaping loop. However, the performance (SNR/ Noise-level) level should remain the same in both cases. This thesis aims at analyzing the implications of using a noise-shaping loop in between the decimation stages of a Δ-Σ ADC and also finding the appropriate decimation filter types that could be used in such a decimation system. This thesis also tries to compare the complexity introduced by using the noise-shaping loop with the reduction achieved in the later decimation stages in terms of the input word length. Filter required in the system will also be optimized using minimax optimization technique.

Visual Object Tracking is the computer vision problem of estimating a target trajectory in a video given only its initial state. A visual tracker often acts as a component in the intelligent vision systems seen in for instance surveillance, autonomous vehicles or robots, and unmanned aerial vehicles. Applications may require robust tracking performance on difficult sequences depicting targets undergoing large changes in appearance, while enforcing a real-time constraint. Discriminative correlation filters have shown promising tracking performance in recent years, and consistently improved state-of-the-art. With the advent of deep learning, new robust deep features have improved tracking performance considerably. However, methods based on discriminative correlation filters learn a rigid template describing the target appearance. This implies an assumption of target rigidity which is not fulfilled in practice. This thesis introduces an approach which integrates deformability into a stateof-the-art tracker. The approach is thoroughly tested on three challenging visual tracking benchmarks, achieving state-of-the-art performance.

It has been aimed to investigate the complexity of discrete-coefficient FIR filters when they are implemented in transposed form and the coefficient redundancy is removed by the n-Dimensional Reduced Adder Graph (RAG-n) approach. Filters with coefficients represented by different quantization schemes have been designed or selected from the literture / their transposed form implemetations after RAG-n process have been compared in terms of complexity. A Genetic Algorithm (GA) based design algorithm has been implemented and used for the design of integer coefficient filters. Algorithms for the realization of filter coefficients in Canonic Signed Digit (CSD) form and realization of n-Dimensional Reduced Adder Graph (RAG-n) have also been implemented. Filter performance is measured as Normalized Peak Ripple Magnitude and implementation complexity as the number of adders used to implement filter coefficients. Number of adders used to implement filter coefficients is calculated by using two different methods: CSD and RAG-n. RAG-n method has been applied to FIR digital filter design methods that don&rsquo / t consider reduction of implementation complexity via RAG-n with transposed direct form filter structure. For implementation complexity, it is concluded that &ldquo / RAG-n algorithm with transposed direct form filter structure&rdquo / provides better results over the &ldquo / CSD, SPT coefficient design followed by transposed direct form filter structure&rdquo / in terms of number of adders used in the implementation.

Ačkoliv můžeme v literatuře nalézt řadu příkladů prezentujících evoluční návrh jakožto zajímavou a slibnou alternativu k tradičním návrhovým technikám používaným v oblasti číslicových obvodů, praktické nasazení je často problematické zejména v důsledku tzv. problému škálovatelnosti, který se projevuje např. tak, že evoluční algoritmus je schopen poskytovat uspokojivé výsledky pouze pro malé instance řešeného problému. Vážný problém představuje tzv. problém škálovatelnosti evaluace fitness funkce, který je markantní zejména v oblasti syntézy kombinačních obvodů, kde doba potřebná pro ohodnocení kandidátního řešení typicky roste exponenciálně se zvyšujícím se počtem primárních vstupů. Tato disertační práce se zabývá návrhem několika metod umožňujících redukovat problem škálovatelnosti evaluace v oblasti evolučního návrhu a optimalizace číslicových systémů. Cílem je pomocí několika případových studií ukázat, že s využitím vhodných akceleračních technik jsou evoluční techniky schopny automaticky navrhovat inovativní/kompetitivní řešení praktických problémů. Aby bylo možné redukovat problém škálovatelnosti v oblasti evolučního návrhu číslicových filtrů, byl navržen doménově specifický akcelerátor na bázi FPGA. Tato problematika reprezentuje případ, kdy je nutné ohodnotit velké množství trénovacích dat a současně provést mnoho generací. Pomocí navrženého akcelerátoru se podařilo objevit efektivní implementace různých nelineárních obrazových filtrů. S využitím evolučně navržených filtrů byl vytvořen robustní nelineární filtr implusního šumu, který je chráněn užitným vzorem. Navržený filtr vykazuje v porovnání s konvenčními řešeními vysokou kvalitu filtrace a nízkou implementační cenu. Spojením evolučního návrhu a technik známých z oblasti formální verifikace se podařilo vytvořit systém umožňující výrazně redukovat problém škálovatelnosti evoluční syntézy kombinačních obvodů na úrovni hradel. Navržená metoda dovoluje produkovat komplexní a přesto kvalitní řešení, která jsou schopna konkurovat komerčním nástrojům pro logickou syntézu. Navržený algoritmus byl experimentálně ověřen na sadě několika benchmarkových obvodů včetně tzv. obtížně syntetizovatelných obvodů, kde dosahoval v průměru o 25% lepších výsledků než dostupné akademické i komerční nástroje. Poslední doménou, kterou se práce zabývá, je akcelerace evolučního návrhu lineárních systémů. Na příkladu evolučního návrhu násobiček s vícenásobnými konstantními koeficienty bylo ukázáno, že čas potřebný k evaluaci kandidátního řešení lze výrazně redukovat (defacto na ohodocení jediného testovacího vektoru), je-li brán v potaz charakter řešeného problému (v tomto případě linearita).

The research reported in this thesis is concerned with robust precoder and transceiver optimization in multiuser multi-antenna wireless communication systems in the presence of imperfect channel state information(CSI). Precoding at the transmit side, which utilizes the CSI, can improve the system performance and simplify the receiver design. Transmit precoding is essential for inter-user interference cancellation in multiuser downlink where users do not cooperate. Linear and non-linear precoding have been widely investigated as low-complexity alternatives to dirty paper coding-based transmission scheme for multiuser multiple-input multiple-output(MU-MIMO)downlink. Similarly, in relay-assisted networks, precoding at the relay nodes have been shown to improve performance. The precoder and joint precoder/receive filter (transceiver) designs usually assume perfect knowledge of the CSI. In practical systems, however, the CSI will be imperfect due to estimation errors, feedback errors and feedback delays. Such imperfections in CSI will lead to deterioration of performance of the precoders/transceivers designed assuming perfect CSI. In such situations, designs which are robust to CSI errors are crucial to realize the potential of multiuser multi-antenna systems in practice. This thesis focuses on the robust designs of precoders and transceivers for MU-MIMO downlink, and for non-regenerative relay networks in the presence of errors in the CSI. We consider a norm-bounded error(NBE) model, and a stochastic error(SE) model for the CSI errors. These models are suitable for commonly encountered errors, and they allow mathematically and computationally tractable formulations for the robust designs. We adopt a statistically robust design in the case of stochastic error, and a minimax or worst-case robust design in the case of norm-bounded error. We have considered the robust precoder and transceiver designs under different performance criteria based on transmit power and quality-of-service(QoS) constraints. The work reported in this thesis can be grouped into three parts, namely,i ) robust linear pre-coder and transceiver designs for multiuser downlink, ii)robust non-linear precoder and transceiver designs for multiuser downlink, and iii)robust precoder designs for non-regenerative relay networks. Linear precoding: In this part, first, a robust precoder for multiuser multiple-input single-output(MU-MISO)downlink that minimizes the total base station(BS)transmit power with constraints on signal-to-interference-plus-noise ratio(SINR) at the user terminals is considered. We show that this problem can be reformulated as a second order cone program(SOCP) with the same order of computational complexity as that of the non-robust precoder design. Next, a robust design of linear transceiver for MU-MIMO downlink is developed. This design is based on the minimization of sum mean square error(SMSE) with a constraint on the total BS transmit power, and assumes that the error in the CSI at the transmitter(CSIT) follows the stochastic error model. For this design, an iterative algorithm based on the associated Karush-Kuhn-Tucker(KKT) conditions is proposed. Our numerical results demonstrate the robust performance of the propose designs. Non-linear precoding: In this part, we consider robust designs of Tomlinson-Harashima precoders(THP) for MU-MISO and MU-MIMO downlinks with different performance criteria and CSI error models. For MU-MISO systems with imperfect CSIT, we investigate the problem of designing robust THPs under MSE and total BS transmit power constraints. The first design is based on the minimization of total BS transmit power under constraints on the MSE at the individual user receivers. We present an iterative procedure to solve this problem, where each iteration involves the solution of a pair of convex optimization problems. The second design is based on the minimization of a stochastic function of the SMSE under a constraint on the total BS transmit power. We solve this problem efficiently by the method of alternating optimization. For MU-MIMO downlink, we propose robust THP transceiver designs that jointly optimize the TH precoder and receiver filters. We consider these transceiver designs under stochastic and norm-bounded error models for CSIT. For the SE model, we propose a minimum SMSE transceiver design. For the NBE model, we propose three robust designs, namely, minimum SMSE design, MSE-constrained design, and MSE-balancing design. Our proposed solutions to these robust design problems are based on iteratively solving a pair of sub-problems, one of which can be solved analytically, and the other can be formulated as a convex optimization problem that can be solved efficiently. Robust precoder designs for non-regenerative relay networks: In this part, we consider robust designs for two scenarios in the case of relay-assisted networks. First, we consider a non-regenerative relay network with a source-destination node pair assisted by multiple relay nodes, where each node is equipped with a single antenna. The set of the cooperating relay nodes can be considered as a distributed antenna array. For this scenario, we present a robust distributed beam former design that minimizes the total relay transmit power with a constraint on the SNR at the destination node. We show that this robust design problem can be reformulated as a semi-definite program (SDP)that can be solved efficiently. Next, we consider a non-regenerative relay network, where a set of source-destination node pairs are assisted by a MIMO-relay node, which is equipped with multiple transmit and multiple receive antennas. For this case, we consider robust designs in the presence of stochastic and norm-bounded CSI errors. We show that these problems can be reformulated as convex optimization problems. In the case of norm-bounded error, we use an approximate expression for the MSE in order to obtain a tractable solution.