Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)

Permanent URI for this collection

https://hdl.handle.net/10889/68

Browse

Now showing 1 - 20 of 740

Open Access
A nonnegative least squares solver for multiple right-hand sides for approximating the nonnegative matrix factorization
(2022-11-04) Κολώνιας, Λεωνίδας; Kolonias, Leonidas
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
Nonnegative Least Squares (NNLS) problems, where the variables are restricted to take only nonnegative values, often arise in many applications and are also at the core of most approaches to solve the nonnegative matrix factorization (NMF), a low-rank matrix approximation problem with nonnegativity constraints. NMF is a data analysis technique used in a great variety of applications such as text mining, image processing, hyperspectral data analysis, computational biology, and clustering. In more detail, the nonnegative factors can be interpreted as data e.g., as images described by pixel intensities or texts represented by vectors of word counts. The mathematical formulation for NMF appears as a non-convex optimization problem, and various types of algorithms have been devised to solve the problem. The first goal of this thesis is to propose a new eﬀicient, yet simple to implement, approach to solve nonnegative linear least squares problems for multiple right-hand sides. More precisely, we study and use properties of global algorithms for least squares problems which are then combined with rules that enforce nonnegativity and lead to novel techniques for solving the aforementioned problem by a flexible Krylov subspace method. Comparisons of the state of the art algorithms using datasets and examples that come from real life applications as well as those artificially generated show that the proposed new algorithm presents a satisfactory behaviour and in some cases outperforms existing ones in computational speed and accuracy. Our second goal is to study extensively the NMF, its properties and applications and dive into the existing algorithms and methodologies used in order to approximate a solution for it. Moreover, using our new approach, tuned to solve large scale nonnegative least squares problems for multiple right-hand sides we present a novel algorithm for NMF based on the alternating nonnegative least squares (ANLS) framework. Extensive experiments on document clustering, images and synthetic datasets indicate the effectiveness of our approach.
Open Access
A novel time-dependent model for route planning in public transit networks
(2021-10-21) Μαχαίρα, Παρασκευή-Μαρία-Μαλεβή; Machaira, Paraskevi-Maria-Malevi
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
In this thesis, the non-FIFO time-dependent model with REalistic vehicle eXchange times (REX) is introduced for schedule-based public transport systems, along with two novel query al- gorithms for computing optimal multimodal journeys for either a single criterion (earliest arrivals) or two criteria (plus number of transfers). The REX model is a considerable en- hancement of the simple time-dependent model, with additional features towards handling non-negligible exchanges from one vehicle to another, as well as supporting non-FIFO in- stances which are typical in public transport, without compromising the space e ciency of the model. Apart from the novelty of the model, REX is also accompanied with two novel query algorithms, whose rationale is some short of “trip-targeted” label-correction propagation process: the rst algorithm, TRIP-based Label-correction propagation Algorithm (TRIPLA), efficiently solves the realistic earliest-arrival routing problem; the second algo- rithm, Multi-criteria TRIP-based Label-correction propagation Algorithm (McTRIPLA), solves the bicriteria variant of the problem, where apart from the earliest-arrival objective, the commuters also care for minimizing the number of vehicle exchanges. The set of Pareto- optimal journeys is discovered when all the arrival-times are bounded. We conduct a thor- ough experimental evaluation on real-world public transit networks which demonstrates that TRIPLA query algorithm for the REX model outperforms signi cantly all state-of-art query algorithms for multimodal routing in schedule-based public transport models, while McTRIPLA is competitive.
Open Access
AI-enhanced perception systems for flexible assembly applications using mobile manipulators
(2022-10-26) Πρέζας, Λουκάς; Prezas, Loukas
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
Assembly is one of the most common but also important processes in today’s manufacturing; thus, the need for efficient automated assembly systems has become an absolute necessity in the recent years. However, automated assembly systems come together with complex challenges such as positioning accuracy, stable gripping, and accommodation of several variants, prevention of error aggregation. Although the use of predefined paths (offline data - waypoints) is a reliable and popular solution, the needs of the Industry 4.0 era for high performance and responsiveness to market demands, impose the use of perception sensors. Perception sensors can increase the robot application robustness regardless of the initial conditions of the operation, enable autonomous decision-making, adaptability, reduction of errors propagation, as well as safe human robot collaboration. This Diploma Thesis focuses on a system implementation, based on vision sensors and Artificial Intelligence so as to create a smart application for high performance parts processing and onsite quality control. The system consists of a robotic manipulator which, equipped with a 3D camera, should use its end-effector for material dispensing and process quality evaluation. The application development is mainly based on OpenCV and PCL library and the communication between the sensors and the robotic manipulator will be achieved with the Robot Operating System (ROS). The initial problem is divided into sub-tasks that the perception system addresses in order to complete the procedure. Specifically, the vision system is responsible for the product recognition (machine learning-based, the correction of the end-effector positioning (defined by offline data) by 3D processing, and the quality inspection. Τhe approach will be tested and validated in a case study inspired for gluing processes in manufacturing.
Open Access
Algorithms and hardware architectures for matrix inversion in massive MIMO uplink data detection
Θάνος, Αλέξιος; Παλιουράς, Βασίλης; Κουφοπαύλου, Οδυσσέας; Μπερμπερίδης, Κωνσταντίνος; Thanos, Alexios
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
This thesis focuses on uplink data detection of a massive MIMO scheme. Two known algorithms for matrix inversion are evaluated considering precision and BER performance for the uplink detection system through MATLAB simulations. Furthermore, exploration of trade-offs in uplink data detection at hardware implementation level and aspects targeting FPGA designs are presented. Design trade-offs include size of datapath units for complexity reduction, hardware architectures for matrix operations, data representation optimization and trading latency for BER performance. Finally, an FPGA-optimized implementation is presented.
Open Access
Algorithms for the fast estimation of statistical leverage scores
Sobczyk, Alexandros; Γαλλόπουλος, Ευστράτιος; Καραγιάννης, Ιωάννης; Ψαράκης, Εμμανουήλ; Σόμπτσυκ, Αλέξανδρος
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
In this thesis we consider algorithms for fast estimations of leverage scores. Statistical leverage scores are a powerful tool for data analysis and statistics and have been successfully used for outlier detection in datasets, locating important nodes in graphs and more recently applied to numerical linear algebra algorithms. In order to build estimators, we consider dimensionality reduction techniques that use randomization in combination with iterative methods for solving linear systems with multiple right hand sides. Based on these techniques we try to overcome certain limitations of the current state-of-the-art algorithms and propose an approach which provably returns good estimations of leverage scores, scales well in parallel/distributed environments and effectively utilizes sparsity. We present our results on synthetic and real world data sets and evaluate its performance, and discuss the advantages and drawbacks relative to all considered approaches.
Open Access
Amorphous parallel algorithms for shortest paths
Παπαδόπουλος, Αναστάσιος; Ζαρολιάγκης, Χρήστος; Ζαρολιάγκης, Χρήστος; Γαλλόπουλος, Ευστράτιος; Κοντογιάννης, Σπύρος; Papadopoulos, Anastasios
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
A new, efficient and highly engineered variant of the well-known Delta-Stepping(DS) algorithm is presented for computing shortest paths in time-dependent networks using amorphous parallelism. This new variant was experimentally evaluated in two scenarios, using real-world data sets (road networks of Berlin, Germany and W. Europe). The experimental study demonstrated that the new DS variant is beneficial for the case of time-dependent road networks and it scales very well with the network size. In particular, the use of the new DS variant for solving the time-dependent many-to-all shortest path problem achieves a speedup of 22.2% compared to the classical time-dependent Dijkstra’s algorithm, while its embedding into state-of-the-art time dependent oracles (that answer in real-time earlier arrival queries for any node pair and departure time) improves the oracle pre-processing phase up to 32% and the oracle query time up to 66%.
Open Access
Application development of a personalized, nutritional type analysis using machine learning
Μπούλιας, Παρασκευάς; Boulias, Paraskevas
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
The following diploma thesis is a research upon topics of diet-related applications and machine learning implementation. It goes into detail explaining past attempts, their specifications, their capabilities, successes and shortcomings. Furthermore, it introduces a modular application of its own with machine learning capabilities as a paradigm to showcase findings and conclusions. A rather unique method of splitting nutrients (e.g. carbs) into arrays is also introduced with the intent of enhancing the system's predictive prowess.
Open Access
Audio-fingerprinting via the k-svd algorthm
Σαραβάνου, Χριστίνα; Saravanou, Christina
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
Music has always played an elemental and unique role in human entertainment and communication. At the end of the 19th century, music was, initially, employed and since then, has been established as a fundamental tool for scientific, medical and educational purposes. In the mid-1960’s, a novel research field, known to most as Music Information Retrieval (MIR)- emerged which aspires to solve various music related problems, such as the song identification and the music genre classification problems, by combing several signal processing and Information Retrieval (IR) techniques. Solving the song identification problem has always been one of the most challenging conundrums of MIR. Throughout the years, several approaches, with the most promising being the audio-fingerprinting scheme, have been proposed to solve this arduous problem. The audio-fingerprinting paradigm, which was introduced in the 1990’s, aims to construct a unique and concise representation -similar to that of a human fingerprint, ergo the name- of an audio track’s signal content. In the last twenty or so years, several alternates of the original audio-fingerprinting scheme have been proposed: Some of which rely on the conventional signal processing and statistical approaches, e.g. the Short-Time Fourier Transform (STFT), while others employ methods and concepts which are applied by several contemporary schemes, such as the Matching Pursuit (MP) algorithm and time-frequency dictionaries-which are used by the Compressive Sensing (CS) and Dictionary Learning (DL) paradigms. This thesis introduces an innovative alternate of the audio-fingerprinting scheme, which aims by employing the Orthogonal Matching Pursuit (OMP) and the K-SVD algorithms -two state-of-the-art techniques applied by the CS and DL paradigms respectively- to construct unique and concise representations of several audio signals to identify their content. Particularly, the suggested approach, initially, aims to create a global dictionary via the K-SVD algorithm and several tens of audio tracks, which uphold the database. The dictionary aspires not only to capture the acoustic/perceptual attributes of the audio signals, but also to apprehend the descriptiveness, the robustness and the discriminability of the audio-fingerprints. Afterwards, the songs-which maintain the database- and the audio excerpt of an unknown audio track- which are used during the querying process-are segmented into several audio frames respectively. Thereupon, the sparse representations of both the audio tracks and clip are computed via the OMP algorithm and the dictionary. Then, the atoms, i.e. the coefficients of an audio signal’s sparse representation, with the highest weight values are extracted and considered equivalent to the most descriptive points of the respective signal’s spectrogram. The landmark pairs, namely four-point peak-pairs, are constructed by using the atoms which were previously selected and are used to construct hash tables. The proposed scheme constructs separate hash tables for every audio track in the database for the audio clip. During the querying process of the suggested paradigm, several voting methods are employed to determine from which song, the audio segment was extracted from and the metadata of the respective song is returned. During the evaluation process of the introduced alternate, several experiments were performed by employing dictionaries of different dimensions. The dictionaries were constructed by using the content of various audio tracks- extracted from two datasets of different size- in both the temporal or the spectral domains. The proposed technique was, initially, assessed to determine which dictionary i.e. which dictionary size and domain, can provide the most accurate results. Moreover, the suggested audio-fingerprinting technique was gauged against its robustness, scilicet whether an audio clip which has been distorted by ambient noise can be identified. Furthermore, the suggested scheme aims to regulate whether an audio clip extracted from a song, which did not partake in the learning process can be identified. In every case, the suggested alternate of the audio-fingerprinting scheme culminated in promising results.
Open Access
Augmenting protein secondary structure information using a novel and efficient computer aided pipeline
Κοροβέσης, Άλκης; Megalooikonomou, Vasileios; Korovesis, Alkis
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
-
Open Access
Autonomous vison-based landing system for unmanned aerial vehicles
Χατζηκαλύμνιος, Ευάγγελος; Chatzikalymnios, Evangelos
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
The use of unmanned aerial vehicles (UAVs) is increasing across many modern civil applica-tion domains, including wireless coverage, delivery, precision agriculture, search and rescue.UAVs equipped with bio-radars, camera sensors and inertial measurement unit (IMU) are life-saving tech-nology that can empower the identification of survivors in catastrophe scenarios and provide medicalaid. However, these UAVs need to be capable of autonomously landing on complex terrains. This isextremely challenging as the structure of these terrains is often unknown, and no prior knowledge canbe leveraged. In this thesis, we present a vision-based autonomous landing system for UAV equippedwith stereo cameras and IMU. The landing site detection algorithm considers several hazardous fac-tors including flatness, steepness and depth accuracy, to compute a weighted cost-map based on whichwe detect dense candidate landing sites. The current pose of theunmanned aerial vehicle (UAV) is es-timated by fusing raw data from the inertial sensors with the pose obtained from stereo ORB-SLAM2.
Open Access
Aνάπτυξη κρυπτογραφικών αλγορίθμων για ετερογενή ασύρματα δίκτυα αισθητήρων
(2012-05-15) Πυργελής, Απόστολος; Σπυράκης, Παύλος; Σταματίου, Ιωάννης; Χατζηγιαννάκης, Ιωάννης; Σπυράκης, Παύλος; Σταματίου, Ιωάννης; Κακλαμάνης, Χρήστος; Pyrgelis, Apostolos
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
Ένα ασύρματο δίκτυο αισθητήρων αποτελείται από χωρικά κατανεμημένες αυτόνομες συσκευές, οι οποίες χρησιμοποιούν αισθητήρες για την συνεργατική παρακολούθηση φυσικών και περιβαλλοντικών συνθηκών. Ένας κόμβος ενός δικτύου αισθητήρων συνήθως αποτελείται από ένα μικροελεγκτή, ένα ραδιο-πομποδέκτη, μια πηγή ενέργειας, η οποία συχνά είναι μια μπαταρία και διαφόρων ειδών αισθητήρες (π.χ. θερμοκρασίας, υγρασίας, φωτεινότητας). Τα ασύρματα δίκτυα αισθητήρων προσφέρουν οικονομικά βιώσιμες λύσεις σε ποικίλες εφαρμογές. Δίκτυα τέτοιου τύπου δραστηριοποιούνται σε βιοιατρικές, στρατιωτικές, βιομηχανικές εφαρμογές καθώς και εφαρμογές παρακολούθησης του φυσικού περιβάλλοντος. Επιπλέον, τα δίκτυα αισθητήρων είναι κλειδί για τη δημιουργία έξυπνων χώρων που ενθέτουν τεχνολογία πληροφορίας στα καθημερινά περιβάλλοντα εργασίας και κατοικίας. Λόγω της χρήσης τέτοιων δικτύων σε περιβάλλοντα που ανταλλάσσονται ευαίσθητες πληροφορίες, δημιουργούνται θέματα ασφάλειας και μυστικότητας. Χαρακτηριστικά των διαφόρων εφαρμογών όπως η λειτουργία σε αντίξοα περιβάλλοντα, η ελλιπής γνώση της τοπολογίας του δικτύου, οι δυνατότητες αυτο-οργάνωσης και αυτόματης διόρθωσης λειτουργιών και η λειτουργία χωρίς ανθρώπινη επιτήρηση καθιστούν τη διατήρηση της ασφάλειας μια μεγάλη πρόκληση. Ένας κλάδος που παρέχει λύσεις σε προβλήματα ασφαλείας είναι αυτός της κρυπτογραφίας. Η κρυπτογραφία είναι μια καλά εδραιωμένη επιστημονική περιοχή, με πρωτόκολλα και πρότυπα τα οποία τυγχάνουν ευρείας αναγνώρισης. Παρόλα αυτά, η χρήση τους σε περιβάλλοντα περιορισμένων πόρων όπως αυτά των ασυρμάτων δικτύων αισθητήρων, απαιτεί προσαρμογές. Η απαίτηση αυτή προκύπτει λόγω των ιδιαίτερων χαρακτηριστικών των δικτύων αυτών και των κόμβων που τα απαρτίζουν όπως η χαμηλή υπολογιστική ισχύς, οι περιορισμένες δυνατότητες αποθήκευσης και η περιορισμένη διαθέσιμη ενέργεια καθώς και η επικοινωνία ασύρματης φύσης που υιοθετείται. Ένα επιπλέον πρόβλημα που παρουσιάζεται στα ασύρματα δίκτυα αισθητήρων, είναι η ετερογένεια. Οι συσκευές αισθητήρων που παράγονται από τη βιομηχανία σήμερα, έχουν διαφορετικές υπολογιστικές δυνατότητες και εκτελούν διαφορετικά λειτουργικά συστήματα. Κάποιες αποτελούνται από μικροεπεξεργαστές 8-bit και έχουν ελάχιστη ποσότητα μνήμης RAM, ενώ άλλες έχουν πολύ μεγάλη υπολογιστική δύναμη και μπορούν να εκτελέσουν desktop λειτουργικά συστήματα όπως Linux. Έτσι, ενώ είναι σχετικά εύκολο να αναπτύξει κανείς μια εφαρμογή για μια συγκεκριμένη πλατφόρμα, είναι πολύ δύσκολο να γράψει γενικό κώδικα ανεξάρτητο της πλατφόρμας μεταγλώτισσης. Έτσι, υπάρχοντες υλοποιημένοι αλγόριθμοι και εφαρμογές πρέπει να τροποποιούνται κατάλληλα για να μπορούν να προσαρμοστούν σε διαφορετικά περιβάλλοντα ανάπτυξης. Μια απάντηση στο παραπάνω πρόβλημα δίνει η βιβλιοθήκη Wiselib η οποία προσφέρει ένα προγραμματιστικό περιβάλλον για την ανάπτυξη γενικών αλγορίθμων που έχουν ως στόχο την εκτέλεσή τους σε ετερογενή δίκτυα αισθητήρων. Σκοπός της παρούσας διπλωματικής εργασίας είναι να προσφέρει λύσεις στα δύο προαναφερθέντα προβλήματα, ανάπτυσσοντας κρυπτογραφικούς αλγόριθμους για ετερογενή ασύρματα δίκτυα αισθητήρων. Για την επίτευξη του σκοπού αυτού, αναπτύσουμε μια κρυπτογραφική βιβλιοθήκη στο προγραμματιστικό περιβάλλον της Wiselib, μιας γενικής βιβλιοθήκης αλγορίθμων για ετερογενή δίκτυα αισθητήρων. Η Wiselib είναι υλοποιημένη σε C++ και με χρήση τεχνικών όπως τα πρότυπα και οι inline συναρτήσεις, επιτρέπει τη συγγραφή γενικού κώδικα ο οποίος αναλύεται και δεσμεύεται κατά τη διαδικασία μεταγλώττισσης χωρίς να δημιουργείται πλεονασμός μνήμης ή υπολογισμού. Λόγω των απαιτήσεων ασφαλείας που δημιουργούνται από τις εφαρμογές δικτύων αισθητήρων καθώς και των περιορισμένων υπολογιστικών πόρων, η κρυπτογραφική μας βιβλιοθήκη παρέχει αλγορίθμους τόσο συμμετρικής όσο και ασυμμετρικής κρυπτογραφίας. Οι αλγόριθμοι ασυμμετρικής κρυπτογραφίας βασίζονται στην κρυπτογραφία ελλειπτικών καμπυλών. Οι ελλειπτικές καμπύλες αποτελούν ένα ιδανικό σύστημα για ανάπτυξη κρυπτογραφίας δημοσίου κλειδιού σε ενσωματωμένα περιβάλλοντα τα οποία υστερούν σε επεξεργαστική ισχύ, μνήμη και ενέργεια. Αυτό ισχύει διότι τα συστήματα ελλειπτικών καμπυλών προσφέρουν το ίδιο επίπεδο ασφάλειας με άλλα κρυπτοσυστήματα (π.χ. RSA) με χρήση πολύ μικρότερου μεγέθους κλειδιών. Έτσι, συνολικά η βιβλιοθήκη μας παρέχει τους εξής αλγορίθμους: τον αλγόριθμο συμμετρικής κρυπτογράφησης AES, τον αλγόριθμο κατακερματισμού SHA-1, το σχήμα συμφωνίας κλειδιών Diffie Hellman (ECDH), τον αλγόριθμο ασυμμετρικής κρυπτογράφησης ECIES και το σχήμα ψηφιακής υπογραφής ECDSA. Για την ανάλυση της απόδοσης της κρυπτογραφικής μας βιβλιοθήκης γίνεται πειραματική αξιολόγηση (χρόνος εκτέλεσης, ενέργεια,μέγεθος μεταφρασμένου κώδικα) των παραπάνω αλγορίθμων σε δύο συσκευές ( iSense, TelosB) με διαφορετικές επεξεργαστικές δυνατότητες (16 MHz, 8 MHz) που τρέχουν διαφορετικά λειτουργικά συστήματα (iSense OS, Contiki Sky). Το γεγονός ότι αξιολογήσαμε τους κρυπτογραφικούς αλγορίθμους σε δύο συσκευές διαφορετικών δυνατοτήτων και περιβαλλόντων ανάπτυξης, αποδεικνύει τη γενικότητα της υλοποίησης μας.Τέλος, για να αποδείξουμε την ευκολία χρήσης των υλοποιημένων αλγορίθμων παρουσιάζουμε τρεις εφαρμογές δικτύων αισθητήρων που τους χρησιμοποιούνε. Πιο συγκεκριμένα, επιδεικνύουμε πως οι κρυπτογραφικοί αλγόριθμοι μπορούν να συνδυαστούν με αλγορίθμους δρομολόγησης και ομαδοποίησης που παρέχει η βιβλιοθήκη Wiselib, με αποτέλεσμα να δημιουργηθούν ασφαλείς εφαρμογές δικτύων αισθητήρων.
Open Access
BCI P300-based speller for control and surveilance
Μαρκόπουλος, Κωνσταντίνος; Markopoulos, Konstantinos
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
In recent years, the development of low-cost, non-invasive and portable electrophysiological systems that record and process brain signals has increased. As a result, Brain Computer Interface (BCI) systems are becoming more accessible to the academic community and the general public, serving different applications and needs, unlike previous years that these systems were much more expensive, more complex in their use, and their application was exclusively in health applications. In this diploma thesis, the goal is to create a BCI system by which a user can control applications and devices through the signals collected from the brain and interact with its environment. At the beginning, the wider field is presented, describing the brain structure, ways of recording information from the brain, BCI categories and the reasons they succeed. Then, the general architectural model of BCI systems is presented. There is information on all the steps that need to be followed in order to create a BCI system as well as many bibliographic references widely used for various kinds of approaches. More specifically, ways of recording the data from the brain according to the targeting of each experiment are being presented, and ways of processing the received signals in order to get rid of noise and strengthen their informational content are being analyzed. There is, also, a bibliographic presentation of methods, that are being presented, by which the features are extracted from the processed signals, aiming at reducing their dimensions and increasing their informational content. Then all this processed data passes through algorithms and machine learning techniques, to produce the final model. In this thesis, a BCI P300-based Speller system is proposed. Emotiv EPOC was the EEG headset used for the experiments. This study aims to describe the design of a real-time EEG-based communication aid system, using brain-computer interface technologies. In more detail, the proposed system consists of a 6x6 matrix display, containing letters and numbers for the spelling procedure. After the spelling is done, the command is driven to a Raspberry PI which connects to all the devices and carries a camera with 2 degrees of freedom combined with computer vision algorithms for the processing. For the speller, an xDAWN spatial filtering is introduced and different classification methods are compared, in order to produce the most accurate and fast system.
Open Access
Behavioral and hardware prototyping of an all digital transmitter
Αποστολίδης, Παναγιώτης; Apostolidis, Panagiotis
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
With the expansion of telecommunications and the arrival of new technologies such as 5G and IoT (Internet of Things) as well as the use of SBCs (Single-board Computers), it has become apparent that small computing nodes need to interact with each other over short distances wirelessly and with low-power consumption. Radio frequencies have been used extensively in many types of wireless communications like mobile communication, Wi-Fi and Bluetooth to name a few. At first, the RF transceiver chain was very analog-intensive. The development of digital communications, allows for some stages to be replaced with their digital implementation. Therefore, the digital baseband signal has superseded its analog counterpart. By introducing this signal to a DAC (Digital-to-Analog Converter), it can be translated in analog baseband, then modulated and amplified to RF frequency bands before being driven to an antenna. Moreover, the merge of digital-to-analog conversion and RF translation in direct digital-to-RF schemes, further reduces the analog components of the transmitter circuitry. Finally, the power amplifier is a high energy consumption component, sometimes adding up to 80\% of the static power consumption of the circuit. Latest designs include switching amplifiers (Class-D amplifiers) that operate as electronic switches rather than linear gain devices in order to reduce the working (ΟΝ) time of the amplifier as much as possible. All the aforementioned design techniques are employed in order to reach a new communication standard with low area, low-power consumption and low noise interference: the all-digital transmitter. This thesis will formulate a behavioral simulation and hardware prototyping for a direct frequency modulation all digital transmitter. The goal is to get performance results pertaining the digital signal output spectrum and phase noise as well as spurious activity. A receiver will also be implemented in order tο have a reference in proper transmission. Ultimately, the transmitter will be synthesized and tested in hardware in an attempt to have a working demo.
Open Access
Block chain για ψηφιακή κυβέρνηση
Παπανικολάου, Γεώργιος; Papanikolaou, George
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
Αυτή η διατριβή ασχολείται με τη χρήση ή την πιθανή χρήση της τεχνολογίας blockchain στην ψηφιακή κυβέρνηση με βάση τη χρήση ψηφιακών υπηρεσιών και ψηφιακών διαδικασιών. Οι συναλλαγές ψηφιακών υπηρεσιών ζητούνται από πολίτες και επιχειρήσεις από κυβερνητικές εφαρμογές συστήματος. Αυτές οι συναλλαγές (αρχεία) αποθηκεύονται σε ψηφιακό ντουλάπι στο οποίο έχει πρόσβαση η επιχείρηση και ο πολίτης ή μεταφορτώνονται στον χώρο αποθήκευσης προσωπικών υπολογιστών. Όταν μια κυβερνητική συναλλαγή (εγγραφή) δημιουργείται και υπογράφεται, μπορεί να γραφτεί στο blockchain, το οποίο παρέχει στους πολίτες και τις επιχειρήσεις την απόδειξη και την εμπιστοσύνη ότι η εγγραφή δεν μπορεί να αλλάξει. Αυτά τα ψηφιακά κυβερνητικά αρχεία μπορούν να κωδικοποιηθούν και να αποθηκευτούν στο blockchain με ένα ιδιωτικό κλειδί, έτσι ώστε να είναι προσβάσιμα μόνο από συγκεκριμένα άτομα, διασφαλίζοντας έτσι το απόρρητο. Το Blockchain, όπως υποδηλώνει το όνομά του, αποτελείται από πολλά μπλοκ που συνδέονται μεταξύ τους. Επιτρέπει τη χρήση μπλοκ για την αποθήκευση νέων δεδομένων και την προσθήκη αυτών των μπλοκ στο blockchain μόλις λάβουν κατακερματισμό. Η τεχνολογία Blockchain επιτρέπει έναν αξιόπιστο και ασφαλή τρόπο αποθήκευσης δεδομένων, οπότε πρακτικά είναι αδύνατο το hacking των μπλοκ δεδομένων. Όταν προστίθεται αυτό το νέο μπλοκ στο blockchain, καθίσταται διαθέσιμο στο κοινό για να το δει ο καθένας. Το Blockchain έχει τη δυνατότητα να εξαλείψει την ανάγκη σάρωσης εγγράφων, να τα αποθηκεύει σε ένα τοπικό σύστημα και να εντοπίζει φυσικά αρχεία σε ένα τοπικό γραφείο εγγραφής ή σε ένα κεντρικό σύστημα. Εάν η ιδιοκτησία ψηφιακών συναλλαγών αποθηκεύεται και επαληθευτεί στο blockchain, οι κάτοχοι μπορούν να εμπιστεύονται ότι η συναλλαγή τους είναι ακριβής και μόνιμη. Ο στόχος αυτής της διατριβής είναι να αναλύσει τις ψηφιακές υπηρεσίες της κυβέρνησης προκειμένου να εφαρμόσει την τεχνολογία blockchain καθώς και τους πιθανούς τρόπους με τους οποίους μπορεί να επιτευχθεί. Για το σκοπό αυτό, θα παρουσιαστούν κατάλληλες περιπτώσεις χρήσης σε μια εφαρμογή blockchain και τα αντίστοιχα αποτελέσματα θα εξαχθούν, θα παρουσιαστούν και θα αναλυθούν. Τέλος, θα αναλυθούν τα πλεονεκτήματα και τα μειονεκτήματα της χρήσης τεχνολογίας blockchain στην ψηφιακή διακυβέρνηση σε σύγκριση με τα κεντρικά συστήματα υπολογιστών.
Open Access
Compression techniques in digital hearing aids
(2009-12-14T10:53:16Z) Γαρίνη, Νικολέτα; Μουστακίδης, Γεώργιος; Μουστακίδης, Γεώργιος; Ψαράκης, Εμμανουήλ; Γαλατσάνος, Νικόλαος; Garini, Nikoleta
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
There has been explosion in the number of digital hearing aids on the market in the last five years. This master thesis deals with some basic issues related to Digital Hearing Aids and more specifically, with the matter of compression in hearing aid devices. The classic frequency-domain compression uses FFT processing and the ideal and practical FFT systems are described. The underlying theory of Multirate Filter Banks and the Polyphase Decomposition as an efficient way of implementing them are presented. A different prototype filter design is thoroughly described and is proposed since it provides a minimum combined approximation error. Moreover, the approximation of the time-domain post filter with gain coefficients being adapted at the frequency domain is done by an all pole filter of lower degree. The simulation results provide us an evaluation of the proposed technique. Our contribution has been the design of a low-delay FIR filter which is extremely crucial for real-time speech processing applications.
Open Access
Coordinated beamforming for hyper-cellular mmWave communications using machine learning
(2023-02-28) Κωνσταντόπουλος, Γεώργιος; Konstantopoulos, Georgios
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
The rapid evolution of technology, the increasing use of wireless devices and the ever-increasing volume of data that needs to be transferred have created the need to design new, innovative standards for the fast and reliable distribution of information. The new generation of 5G wireless telecommunications is set to provide a solution to this problem. The 5G generation promises a big increase in data transmission speeds, as well as global coverage through the interconnection of all devices in a network and a large increase in network coverage stations. It is understood that managing such a large network is a complex process and requires large amounts of energy to achieve. Advances in machine learning are creating new perspectives for the design of fifth generation telecommunication systems and for the optimization of automated data management techniques. The purpose of this paper is to present an innovative concept in which machine learning techniques can be used to select the antenna stations that will serve each user and how they will be served, in a mmWave coordinated beamforming scenario. The consequence of this technique is a drastic reduction of the energy footprint of the network, through the temporary deactivation of antenna stations not selected to serve users. It has to be noted that for our simulations we will use the system model and the data of the DeepMIMO project. The present work contains six chapters. Chapter 1 provides a detailed description of the fifth generation systems and the technologies they use. Chapter 2 discusses hyper-cellular networks which is a specific type of network we will deal with and gives basic concepts and information about it. Chapter 3 gives an extensive description of the subject of machine learning and in particular neural networks and ways of training them. Chapter 4 describes the implementation of the DeepMIMO system model and discusses in detail the concept of coordinated beamforming. In Chapter 5 there is the final implementation of the system and its simulation results and in Chapter 6 the main conclusions of the paper are collected and suggestions for further research of the system are made.
Open Access
Data representation and adder architectures in the presence of variations
Παπαχατζόπουλος, Κλεάνθης; Παλιουράς, Βασίλειος; Θεοδωρίδης, Γεώργιος; Νικολός, Δημήτριος; Papachatzopoulos, Kleanthis
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
It is well known that reduced logic depth allows for operation at low voltages, therefore reducing power dissipation. However, such circuits are particularly susceptible to variations, which may compromise expected benefits. This Master Thesis is focused on the evaluation of the performance and power character- istics of certain adder structures under inter- and intra-die process variations in deep-submicron technology nodes. Specifically, it presents a solution for low-power addition under variability, which successfully handles the challenge of increased threshold voltage variation. We quanti- tatively compare the impact of variation on the performance of Ripple-Carry Adder (RCA) and Borrow-Save Adder (BSA), and quantify the average power reduction achieved by BSA attained at low voltage values, at the cost of increased delay variation. In addition, we propose a tech- nique that enhances BSA tolerance to variations. Using Statistical SPICE Timing Evaluation at 45-nm, 32-nm, and 16-nm nodes, we estimate the maximum critical path delay variation and average power dissipation of BSA at different supply voltages. Our analysis reveals that BSA achieves three times smaller standard deviation of maximum delay than RCA at the same supply voltage for a 45-nm technology node. In addition, we show that it is possible to substantially reduce the supply voltage, decreasing by almost 60% the overall power dissipation of BSA in comparison to a counterpart operating at nominal voltage, while keeping maximum delay less than that of RCA. Furthermore, simple design optimizations in the design of BSA are intro- duced that trade latency for variability, significantly reducing normalized standard deviation of the maximum delay. BSIM 4 MOSFET libraries have been employed for a precise evaluation of performance-power characteristics in todays technology nodes. Furthermore, this Master Thesis introduces two statistical delay-variability models for RCA and BSA. The models consider both intra- and inter-die delay variations. The first proposed model, named as Type-I model, is derived in the form of expressions for the computation of the exact Probability Density Functions (PDFs) of maximum output delays of the two adder archi- tectures. Furthermore, closed formulas for the correlation coefficients between output delays of the aforementioned adder architectures are presented. The introduced derived correlation coef- ficients are subsequently combined with Clark’s method to derive the second proposed model, Type-II model, which comprises approximations of the maximum delay PDFs of RCA and BSA. The proposed Clark-based Type-II model uses Gaussian distributions to approximate maximum delay distributions, taking into account the correlation between logic paths. Simulation results and the derived exact Type-I PDFs are found to perfectly agree, while the proposed Clark-based Type-II models present an error for standard deviation of maximum delay that increases as BSA word length increases. Both the introduced models and the simulations prove that BSAs achieve narrower delay distributions than RCAs, i.e., they significantly reduce delay variance. Conse- quently, BSAs are proven to be suitable for variation-tolerant applications by providing a timing safety margin, when compared to RCA architectures. The underlying analysis indicates that, for the case of BSA and (inter- and) intra-die delay variations, the Type-II models introduce no-negligible errors, which are as much as 16% of the standard deviation of maximum delay for a 256-bit BSA, as the Type-II Gaussian PDF approximations deviate significantly from the exact Type-I PDFs. However, for all RCA and BSA inter-die only variation cases, both Types present satisfactory accuracy due to Gaussian shape of exact PDFs.
Open Access
Data representation and hardware aspects in a fully-folded successive cancellation polar decoder
Ανδριακόπουλος, Χρήστος; Παλιουράς, Βασίλειος; Παλιουράς, Βασίλειος; Αντωνακόπουλος, Θεόδωρος; Μπερμπερίδης, Κωνσταντίνος; Andriakopoulos, Chris
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
In this master thesis, is studied and implemented Error-Correction Codes system based on Polar Codes. Polar Codes, introduced by Arıkan, are known for the capacity approach- ing with low hardware complexity. We simulated a Polar Encoder/Decoder system on Matlab in order to evaluate Polar Codes, and implemented in a FPGA device, afterwards. At the beginning, is introduced the basic model of a digital communication system, ba- sic units that consist of the entire communication system. After that, is introduced Polar Codes, the construction procedure, and how can be implemented Polar Encoder and De- coder in hardware. Chapter. 3 introduces decoding algorithms for Polar Codes, such as Successive-Cancellation (SC), Successive-List-Cancellation (SLC) and Belief Propaga- tion (BP). Also Chapter 3 discusses several state-of-the-art implementations of Polar De- coders. In this master thesis, we focus on SC decoding algorithm due its simplicity which facilitate the study of codes with relatively long block length. Due to limited resources in a FPGA device, the technique of ‘folding’ applied, so the transformed DFG uses only one processing unit (PB). A data representation scheme is introduced and it is shown to lead to 30−50% reduction of required memory, for practical block lengths. In Chapter 5 shows implementation results and BER vs. noise level curves, while data are extracted from Matlab simulations and FPGA implementation using the logic analyzer Chipscope.
Open Access
Deep learning based image compression
Ηλιοπούλου, Σοφία; Iliopoulou, Sofia
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
Image compression is a research topic that has interested both the academic community and the business world for decades. The large size of media files combined with the popularity of social media and streaming services, render image compression necessary. However, it is imperative to maintain a decent image quality while reducing its size. Many methods have been developed and are still used, but in the last few years Artificial Intelligence has been making remarkable progress in the field. More and more image compression techniques choose to utilize Deep Learning now. This thesis first illustrates the range of application of Artificial Intelligence and then presents the related work in the field of Deep-Learning-based image compression. However, it is important to also study the traditional codecs like JPEG, since they are still used to this day, and they have become benchmarks for every compression algorithm. To this effect, the traditional image compression techniques are presented. Specifically, multiple methods are briefly explained, based on their effect on image quality. In addition, the different types of neural networks that are used for processing and compressing images, are presented. Their architecture and use are explained. Then, each part of the proposed implementation for image dimensionality reduction is shown. Finally, the performance of the developed system is evaluated and compared with other relevant attempts. In conclusion, the results show that the implementation of a Deep-Learning-based image compression method that is equal or even surpasses the traditional techniques in efficiency, is possible and high performance can be achieved. In a future work, the effect of different types of neural networks, as well as the improvement of the current method’s results will be studied.
Open Access
Deep-learning-based forward-error-correction decoding techniques and optimizations for hardware implementation
Καββουσανός, Εμμανουήλ; Παλιουράς, Βασίλης; Paliouras, Vassilis; Kavvousanos, Emmanouil
Τμήμα Μηχανικών Η/Υ και Πληροφορικής (ΜΔΕ)
In recent years, Deep-Learning has been adopted by a wide spectrum of applications, as it is a powerful problem-solving methodology which can be applied in extremely diverse fields. Various types of Artificial Neural Networks can be trained to perform a task with high accuracy. The effectiveness of such networks can surpass even humans in computer vision and language processing problems. Beyond the typical aforementioned applications, Deep-Learning techniques have been recently examined for adoption in several Telecommunication areas, including Forward Error Correction. Several works have investigated the training of neural networks for channel decoding. In this thesis, the case of the Syndrome-based Deep-Learning Decoder is considered for the BCH(63,45) code and transmission with BPSK modulation through an AWGN channel. First of all, the training process of the neural network decoder is examined, by searching for the optimal training hyperparameters. Furthermore, new neural network decoder architectures are explored, beyond those suggested in the literature and modifications to the existing decoding framework are suggested which improve the decoding performance remarkably. Moreover, the computational complexity of the Syndrome-based DL decoder is considered. Deep-Learning decoding methods are hard to implement in hardware as they normally require millions of operations for inference. In order for Deep-Learning decoding to be a competitive candidate for practical applications, further research effort is required to reduce the computational complexity and storage requirements of the Neural Networks involved. In this thesis, a structured flow is presented that significantly compresses a trained Syndrome-Based Neural Network Decoder by pruning up to 80% of the network weights and quantizing them to 8-bit fixed-point representation. The attained compressed Neural Network can then be used for inference, by designing special hardware or by using a generic Deep-Learning hardware accelerator that exploits the compressed structure of the network. Finally, the deployment of the DL Decoder in an embedded application is showcased, using the AI Edge platform by Xilinx. Implementation results are provided for the compressed DL Decoder, regarding latency, throughput rate and BER performance.