The move from Standard Definition (SD) to High Definition (HD) represents a six times increases in data, which needs to be processed. With expanding resolutions and evolving compression, there is a need for high performance with flexible architectures to allow for quick upgrade ability. The technology advances in image display resolutions, advanced compression techniques, and video intelligence. Software implementation of these systems can attain accuracy with tradeoffs among processing performance (to achieve specified frame rates, working on large image data sets), power and cost constraints. There is a need for new architectures to be in pace with the fast innovations in video and imaging. It contains dedicated hardware implementation of the pixel and frame rate processes on Field Programmable Gate Array (FPGA) to achieve the real-time performance. The following outlines the contributions of the dissertation. (1) We develop a target detection system by applying a novel running average mean threshold (RAMT) approach to globalize the threshold required for background subtraction. This approach adapts the threshold automatically to different environments (indoor and outdoor) and different targets (humans and vehicles). For low power consumption and better performance, we design the complete system on FPGA. (2) We introduce a safe distance factor and develop an algorithm for occlusion occurrence detection during target tracking. A novel mean-threshold is calculated by motion-position analysis. (3) A new strategy for gesture recognition is developed using Combinational Neural Networks (CNN) based on a tree structure. Analysis of the method is done on American Sign Language (ASL) gestures. We introduce novel point of interests approach to reduce the feature vector size and gradient threshold approach for accurate classification. (4) We design a gesture recognition system using a hardware/ software co-simulation neural network for high speed and low memory storage requirements provided by the FPGA. We develop an innovative maximum distant algorithm which uses only 0.39% of the image as the feature vector to train and test the system design. Database set gestures involved in different applications may vary. Therefore, it is highly essential to keep the feature vector as low as possible while maintaining the same accuracy and performance

Current exercise machines create resistance to motion and dissipate energy as heat. Some companies create ways to harness this energy, but not cost-effectively. The Energy Harvesting from Exercise Machines (EHFEM) project reduces the cost of harnessing the renewable energy. The system architecture includes the elliptical exercise machines outputting power to DC-DC converters, which then connects to the microinverters. All microinverter outputs tie together and then connect to the grid. The control system, placed around the DC-DC converters, quickly detects changes in current, and limits the current to prevent the DC-DC converters and microinverters from entering failure states. An artificial neural network learns to mitigate incohesive microinverter and DC-DC converter actions. The DC-DC converter outputs 36 V DC operating within its specifications, but the microinverter drops input resistance looking for the sharp decrease in power that a solar panel exhibits. Since the DC-DC converter behaves according to Ohm’s Law, the inverter sees no decrease in power until the voltage drops below the microinverter’s minimum input voltage. Once the microinverter turns off, the converter regulates as intended and turns the microinverter back on only to repeat this detrimental cycle. Training the neural network with the back propagation algorithm outputs a value corresponding to the feedback voltage, which increases or decreases the voltage applied from the resistive feedback in the DC-DC converter. In order for the system to react well to changes on the order of tens of microseconds, it must read ADC values and compute the output neuron value quicker than previous control attempts. Measured voltages and currents entering and leaving the DC-DC converter constitute the neural network’s input neurons. Current and voltage sensing circuit designs include low-pass filtering to reduce software noise filtering in the interest of speed. The complete solution slightly reduces the efficiency of the system under a constant load due to additional component power dissipation, while actually increasing it under the expected varying loads.

This master's thesis describes the design of effective working artificial neural network in FPGA Virtex-5 series with the maximum use of the possibility of parallelization. The theoretical part contains basic information on artificial neural networks, FPGA and VHDL. The practical part describes the used format of the variables, creating non-linear function, the principle of calculation the single layers, or the possibility of parameter settings generated artificial neural networks.

This work focuses on the theory of artificial neural networks: the history, individual ways of learning and architecture of networks. It is also necessary to desribe the image processing blocks from scanning and image processing through segmentation to object recognition. The next part is focused on connecting the previous two parts, and therefore on the use of neural networks in image processing, specifically the identification of objects. In the practical part of the work is designed the user application for recognizing characters such as numbers, uppercase and lowercase letters.

Remote Electrical Tilt optimization is an effective method to obtain the optimal Key Performance Indicators (KPIs) by remotely controlling the base station antenna’s vertical tilt. To improve the KPIs aims to improve antennas’ cooperation effect since KPIs measure the quality of cooperation between the antenna to be optimized and its neighbor antennas. Reinforcement Learning (RL) is an appropriate method to learn an antenna tilt control policy since the agent in RL can generate the optimal epsilon greedy tilt optimization policy by observing the environment and learning from the state- action pairs. However, existing models only produced tilt modification strategies by interpreting the to- be- optimized antenna’s features, which cannot fully characterize the mobile cellular network formed by the to- be- optimized antenna and its neighbors. Therefore, incorporating the features of the neighboring antennas into the model is an important measure to improve the optimization strategy. This work will introduce the Graph Attention Network to model the neighborhood antenna’s impact on the antenna to be optimized through the attention mechanism. Furthermore, it will generate a low- dimensional embedding vector with more expressive power to represent the to- be- optimized antenna’s state in the RL framework through dealing with graph- structural data. This new model, namely Graph Attention Q- Network (GAQ), is a model based on DQN and aims to acquire a higher performance than the Deep Q- Network (DQN) model, which is the baseline, evaluated by the same metric — KPI Improvement. Since GAQ has a richer perception of the environment than the vanilla DQN model, it thereby outperforms the DQN model, obtaining fourteen percent performance improvement compared to the baseline. Besides, GAQ also performs 14 per cent better than DQN in terms of convergence efficiency. / Optimering av fjärrlutning är en effektiv metod för att nå optimala nyckeltal genom fjärrstyrning av den vertikala lutningen av en antenn i en basstation. Att förbättra nyckeltalen innebär att förbättra sammarbetseffekten mellan antenner eftersom nyckeltalen är mått på kvalitén av sammarbetet mellan den antenn som optimeras och dess angränsande antenner. Förstärkande Inlärning (FI) är en lämplig metod för att lära sig en optimal strategi för reglering av antennlutningen eftersom agenten inom FI kan generera den optimala epsilongiriga optimeringsstrategin genom att observera miljön och lära sig från par av tillstånd och aktioner. Nuvarande modeller genererar dock endast lutningsstrategier genom att tolka egenskaperna hos den antenn som ska optimeras, vilket inte är tillräckligt för att karatärisera mobilnätverket bestående av antennen som ska optimeras samt dess angränsande antenner. Därav är inkluderingen av de angränsande antennernas egenskaper i modellen viktig för att förbättra optimeringsstrategin. Detta arbete introducerar Graf- Uppmärksammat Nätverk för att modellera de angränsande antennernas påverkan på den antenn som ska optimeras genom uppmärksamhetsmekanismen. Metoden genererar en lågdimensionell vektor med större förmåga att representera den optimerade antennens tillstånd i FI modellen genom att hantera data i struktur av en graf. Den nya modellen, Graf- Uppmärksammat Q- Nätverk (GUQ), är en modell baserad på DQN med mål att nå bättre prestanda än en standard DQN- modell, utvärderat efter samma mätvärde –– förbättring av nyckeltalen. Eftersom GUQ har en större upfattning av miljön så överträffar metoden DQN- modellen genom en fjorton procent bättre prestandaökning. Dessutom, så överträffar GUQ även DQN i form av snabbare konvergens.

Master’s thesis describes adaptive optimal controller design which change parameters of algorithm based on the system information regard for optimal criterion. Generally, the optimal controller solves the problem of minimum states vector. Problems of desired value and steady-state error are solved by variation in optimization algorithm.

Text in Afrikaans / Summaries in Afrikaans and English / In Afrikaans, soos in NederJands en Duits, word saamgestelde woorde aanmekaar geskryf. Nuwe woorde word dus voortdurend geskep deur woorde aanmekaar te haak Dit bemoeilik die proses van woordafkapping tydens teksprosessering, wat deesdae deur rekenaars gedoen word, aangesien die verwysingsbron gedurig verander. Daar bestaan verskeie afkappingsalgoritmes en tegnieke, maar die resultate is onbevredigend. Afrikaanse woorde met korrekte lettergreepverdeling is net die elektroniese weergawe van die handwoordeboek van die Afrikaanse Taal (HAT) onttrek. 'n Neutrale netwerk ( vorentoevoer-terugpropagering) is met sowat. 5 000 van hierdie woorde afgerig. Die neurale netwerk is verfyn deur 'n gcskikte afrigtingsalgoritme en oorfragfunksie vir die probleem asook die optimale aantal verborge lae en aantal neurone in elke laag te bepaal. Die neurale netwerk is met 5 000 nuwe woorde getoets en dit het 97,56% van moontlike posisies korrek as of geldige of ongeldige afkappingsposisies geklassifiseer. Verder is 510 woorde uit tydskrifartikels met die neurale netwerk getoets en 98,75% van moontlike posisies is korrek geklassifiseer. / In Afrikaans, like in Dutch and German, compound words are written as one word. New words are therefore created by simply joining words. Word hyphenation during typesetting by computer is a problem, because the source of reference changes all the time. Several algorithms and techniques for hyphenation exist, but results are not satisfactory. Afrikaans words with correct syllabification were extracted from the electronic version of the Handwoordeboek van die Afrikaans Taal (HAT). A neural network (feedforward backpropagation) was trained with about 5 000 of these words. The neural network was refined by heuristically finding a suitable training algorithm and transfer function for the problem as well as determining the optimal number of layers and number of neurons in each layer. The neural network was tested with 5 000 words not the training data. It classified 97,56% of possible points in these words correctly as either valid or invalid hyphenation points. Furthermore, 510 words from articles in a magazine were tested with the neural network and 98,75% of possible positions were classified correctly. / Computing / M.Sc. (Operasionele Navorsing)

在全球經貿體系自由化下，國際資金流通快速，匯率變動也非常頻繁，廠商的產銷決策與營運，面對匯率風險更加難以掌控。如何掌握匯率的變動，並採取有效的避險措施，是廠商從事貿易必須面臨之重要課題。本研究採用自我迴歸整合移動平均模式、倒傳遞類神經網路及混合式自我迴歸整合移動平均模式及倒傳遞類神經網路模型進行未來即期匯率報酬率之預測。試圖找出合適的新台幣兌美元即期匯率之預測模型，並將其應用於外匯避險操作。 研究結果顯示，關於預測誤差的績效表現，整體來說，以自我迴歸整合移動平均及倒傳遞類神經網路混合式模型表現最佳，顯示傳統時間序列模型捕捉匯率報酬率走勢之能力，藉由倒傳遞類神經網路捕捉其線性預測誤差中非線性的部分，可更符合資料的特性，加強匯率報酬率預測的準確性。考慮預測方向的正確性，在兩個不同的準則下(SR、PT)，皆以自我迴歸整合移動平均模型表現最差，代表其在進行匯率報酬率之預測時正確率較為不足。而在PT檢定當中，倒傳遞類神經網路模型及混合式模型皆達到顯著。因此利用人工智慧模型對報酬率之方向進行預測是有效的，又以自我迴歸整合移動平均及倒傳遞類神經網路混合式模型表現最好。總結來說，利用倒傳遞類神經網路模型針對自我迴歸整合移動平均模型做非線性的調整，同時涵蓋未來匯率報酬率線性與非線性的部分，使得自我迴歸整合移動平均模型之預測誤差、方向準確性皆得到改善，藉由倒傳遞類神經網路捕捉其線性預測誤差中非線性的部分，可更符合資料的特性，加強匯率報酬率預測的準確性。

Text in Afrikaans / Summaries in Afrikaans and English / In Afrikaans, soos in NederJands en Duits, word saamgestelde woorde aanmekaar geskryf. Nuwe woorde word dus voortdurend geskep deur woorde aanmekaar te haak Dit bemoeilik die proses van woordafkapping tydens teksprosessering, wat deesdae deur rekenaars gedoen word, aangesien die verwysingsbron gedurig verander. Daar bestaan verskeie afkappingsalgoritmes en tegnieke, maar die resultate is onbevredigend. Afrikaanse woorde met korrekte lettergreepverdeling is net die elektroniese weergawe van die handwoordeboek van die Afrikaanse Taal (HAT) onttrek. 'n Neutrale netwerk ( vorentoevoer-terugpropagering) is met sowat. 5 000 van hierdie woorde afgerig. Die neurale netwerk is verfyn deur 'n gcskikte afrigtingsalgoritme en oorfragfunksie vir die probleem asook die optimale aantal verborge lae en aantal neurone in elke laag te bepaal. Die neurale netwerk is met 5 000 nuwe woorde getoets en dit het 97,56% van moontlike posisies korrek as of geldige of ongeldige afkappingsposisies geklassifiseer. Verder is 510 woorde uit tydskrifartikels met die neurale netwerk getoets en 98,75% van moontlike posisies is korrek geklassifiseer. / In Afrikaans, like in Dutch and German, compound words are written as one word. New words are therefore created by simply joining words. Word hyphenation during typesetting by computer is a problem, because the source of reference changes all the time. Several algorithms and techniques for hyphenation exist, but results are not satisfactory. Afrikaans words with correct syllabification were extracted from the electronic version of the Handwoordeboek van die Afrikaans Taal (HAT). A neural network (feedforward backpropagation) was trained with about 5 000 of these words. The neural network was refined by heuristically finding a suitable training algorithm and transfer function for the problem as well as determining the optimal number of layers and number of neurons in each layer. The neural network was tested with 5 000 words not the training data. It classified 97,56% of possible points in these words correctly as either valid or invalid hyphenation points. Furthermore, 510 words from articles in a magazine were tested with the neural network and 98,75% of possible positions were classified correctly. / Computing / M.Sc. (Operasionele Navorsing)

L'échographie est aujourd'hui l'une des modalités les plus populaires de diagnostic médical. Il permet d'observer, en temps réel, le mouvement des organes qui facilite le diagnostic des pathologies pour des médecins. L'échocardiographie [1, 2], l'imagerie du flux sanguin [3, 4] et l’élastographie [5-7] sont les domaines préférés de l'estimation de mouvement en utilisant l'échographie (en raison de son haut frame-rate).En conséquence, les images avec meilleurs qualités sont nécessaires. . En imagerie cardiaque, le système classique d'imagerie est limité dans la direction transversale (la direction perpendiculaire à celle de propagation). Travaillant sur la formation des images, ce problème peut être résolu en modifiant la façon de formateur de voie afin d'introduire des oscillations transversales (OTs) dans la fonction d’étalement du point (PSF). La technique d’oscillation transversale a montré son potentiel d'améliorer la précision de l'estimation de mouvement local dans la direction transversale (la direction perpendiculaire à celle de propagation). La classique OT en géométrie linéaire, basée sur l'approximation de Fraunhofer, relie la PSF et la fonction de pondération par la transformée de Fourier. Motivé par l'adaptation des OTs en échocardiographie, nous proposons une technique spécifique basée sur la rétro-propagation afin de construire des OTs en géométrie sectorielle. La performance de la méthode de rétro-propagation proposée a été étudiée progressivement, comparée avec la méthode de la transformée de Fourier, par exemple, l'évaluation de la qualité de la PSF quantifié, dans l'estimation de mouvement cardiaque en simulation, et en étude la qualité des PSF visuellement expérimentale. Les résultats quantifiés montrent les OT-images sont mieux contrôlés par la méthode proposée que par le formateur de voie conventionnelle. Une autre méthode, basée sur la décomposition d'onde plane et un principe différent de rétro-propagation, a été présentée. Cette méthode mieux prend en compte la propriété 2D de PSF, en décomposant la PSF dans un ensemble d'ondes planes directionnelle, les rétro-propage à la sonde, en utilisant les résultats de superposition comme excitations, un PSF simulée et conforme fortement au PSF théorique est acquis. En adaptant cette méthode à la géométrie sectorielle, la qualité de la PSF obtenue en face et sur la côté de la sonde est meilleure en utilisant la décomposition en ondes planes à celle de la transformée de Fourier, le travail supplémentaire sera adressé à adapter la décomposition en ondes planes à imagerie sectorielle et l’estimation du mouvement. / Echography is nowadays one of the most popular medical diagnosis modalities. It enables real-time observation the motion of moving organs which facilitates the diagnosis of pathologies for physician. Echocardiography [1, 2], blood flow imaging [3, 4] and elastography [5-7] are the favorite domains of motion estimation in using of echography (e.g., due to its high frame-rate capacity). Thus the requirements for imaging with high quality are on the primary place. In cardiac imaging, the conventional imaging system is somehow limited in the transverse direction (the direction perpendicular to the beam axis). Working on the image formation, this problem can be addressed by modifying the beamforming scheme in order to introduce transverse oscillations (TOs) in the system point spread function (PSF). Transverse oscillation techniques have shown their potential for improving the accuracy of local motion estimation in the transverse direction (i.e., the direction perpendicular to the beam axis). The conventional design of TOs in linear geometry, which is based on the Fraunhofer approximation, relates PSF and apodization function through a Fourier transform. Motivated by the adaptation of TOs in echocardiography, we propose a specific beamforming approach based on back-propagation in order to build TOs in sectorial geometry. The performance of the proposed back-propagation method has been studied gradually, in comparison with the Fourier transform, such as in evaluation of the quality of PSF, in estimation of simulated cardiac motion and in experiments study, etc. The quantified results demonstrate the proposed method leads to better controlled TOs images than the conventional beamforming. Another method based on plane wave decomposition and a different back-propagation principle has been presented. This method is better taking into account the 2D property of PSF, by decomposing the PSF into a set of plane waves directionally, back-propagating them to the probe, by using the superposition results as excitations, a simulated PSF with high accordance to the theoretical one is acquired. By adapting this method to sectorial geometry, the quality of PSF obtained in front of probe is better using the plane wave decomposition method than that of Fourier relation, but it is limited for the scanning on the side of probe, so the further work will be addressed to adapting the plane wave decomposition method to the complete sectorial imaging.