We study the loss of information (measured in terms of the Kullback- Leibler distance) caused by observing "grouped" data (observing only a discretized version of a continuous random variable). We analyse the asymptotical behaviour of the loss of information as the partition becomes finer. In the case of a univariate observation, we compute the optimal rate of convergence and characterize asymptotically optimal partitions (into intervals). In the multivariate case we derive the asymptotically optimal regular sequences of partitions. Forthermore, we compute the asymptotically optimal transformation of the data, when a sequence of partitions is given. Examples demonstrate the efficiency of the suggested discretizing strategy even for few intervals. (author's abstract) / Series: Forschungsberichte / Institut für Statistik

To push the envelope of DCT-based lossy image/video compression, this thesis is motivated to revisit design of some fundamental blocks in image/video coding, ranging from source modelling, quantization table, quantizers, to entropy coding. Firstly, to better handle the heavy tail phenomenon commonly seen in DCT coefficients, a new model dubbed transparent composite model (TCM) is developed and justified. Given a sequence of DCT coefficients, the TCM first separates the tail from the main body of the sequence, and then uses a uniform distribution to model DCT coefficients in the heavy tail, while using a parametric distribution to model DCT coefficients in the main body. The separation boundary and other distribution parameters are estimated online via maximum likelihood (ML) estimation. Efficient online algorithms are proposed for parameter estimation and their convergence is also proved. When the parametric distribution is truncated Laplacian, the resulting TCM dubbed Laplacian TCM (LPTCM) not only achieves superior modeling accuracy with low estimation complexity, but also has a good capability of nonlinear data reduction by identifying and separating a DCT coefficient in the heavy tail (referred to as an outlier) from a DCT coefficient in the main body (referred to as an inlier). This in turn opens up opportunities for it to be used in DCT-based image compression. Secondly, quantization table design is revisited for image/video coding where soft decision quantization (SDQ) is considered. Unlike conventional approaches where quantization table design is bundled with a specific encoding method, we assume optimal SDQ encoding and design a quantization table for the purpose of reconstruction. Under this assumption, we model transform coefficients across different frequencies as independently distributed random sources and apply the Shannon lower bound to approximate the rate distortion function of each source. We then show that a quantization table can be optimized in a way that the resulting distortion complies with certain behavior, yielding the so-called optimal distortion profile scheme (OptD). Guided by this new theoretical result, we present an efficient statistical-model-based algorithm using the Laplacian model to design quantization tables for DCT-based image compression. When applied to standard JPEG encoding, it provides more than 1.5 dB performance gain (in PSNR), with almost no extra burden on complexity. Compared with the state-of-the-art JPEG quantization table optimizer, the proposed algorithm offers an average 0.5 dB gain with computational complexity reduced by a factor of more than 2000 when SDQ is off, and a 0.1 dB performance gain or more with 85% of the complexity reduced when SDQ is on. Thirdly, based on the LPTCM and OptD, we further propose an efficient non-predictive DCT-based image compression system, where the quantizers and entropy coding are completely re-designed, and the relative SDQ algorithm is also developed. The proposed system achieves overall coding results that are among the best and similar to those of H.264 or HEVC intra (predictive) coding, in terms of rate vs visual quality. On the other hand, in terms of rate vs objective quality, it significantly outperforms baseline JPEG by more than 4.3 dB on average, with a moderate increase on complexity, and ECEB, the state-of-the-art non-predictive image coding, by 0.75 dB when SDQ is off, with the same level of computational complexity, and by 1 dB when SDQ is on, at the cost of extra complexity. In comparison with H.264 intra coding, our system provides an overall 0.4 dB gain or so, with dramatically reduced computational complexity. It offers comparable or even better coding performance than HEVC intra coding in the high-rate region or for complicated images, but with only less than 5% of the encoding complexity of the latter. In addition, our proposed DCT-based image compression system also offers a multiresolution capability, which, together with its comparatively high coding efficiency and low complexity, makes it a good alternative for real-time image processing applications.

This thesis explores the study and design of an all-digital VCO-based ADC in a 65 nm CMOS technology. As the CMOS process enters the deep submicron region, the voltage-domain-based ADCs begins to suffer in improving their performance due to the use of complex analog components. A promising solution to improve the performance of an ADC is to employ as many as possible digital components in a time-domain-based ADC, where it uses the time resolution of an analog signal rather than the voltage resolution. In comparison, as the CMOS process scales down, the time resolution of an analog signal has found superior than the voltage resolution of an analog signal. In recent years, such time-domain-based ADCs have been taken an immense interest due to its inherent features and their design reasons. In this thesis work, the VCO-based ADC design, falls under the category of time-based ADCs which consists of a VCO and an appropriate digital processing circuitry. The employed VCO is used to convert a voltage-based signal into a time signal and thereby it also acts as a time-based quantizer. Then the resulting quantized-time signal is converted into a digital signal by an appropriate digital technique. After different architecture exploration, a conventional VCO-based ADC architecture is implemented in a high-level model to understand the characteristic behaviour of this time-based ADC and then a comprehensive functional schematic-level is designed in reference with the implemented behavioural model using cadence design environment. The performance has been verified using the mixed-levels, of transistor and behavioural-levels due to the greater simulation time of the implemented design. ADC’s dynamic performance has been evaluated using various experiments and simulations. Overall, the simulation experiments showed that the design was found to reach an ENOB of 4.9-bit at 572 MHz speed of sample per second, when a 120 MHz analog signal is applied. The achieved peak performance of the design was a SNR of 40 dB, SFDR of 34 dB and an SNDR of 31 dB over a 120 MHz BW at a 1 V supply voltage. Without any complex building blocks, this VCO-based all-digital ADC design provided a key feature of inherent noise shaping property and also found to be well compatible at the deep submicron region.

Adaptive Resource Allocating Vector Quantizer (ARAVQ) är en teknik för datareducering för mobila robotar. Tekniken har visats framgångsrik i enkla miljöer och det har spekulerats i att den kan fungera som ett generellt datautvinningsverktyg för tidsserier. I rapporten presenteras experiment där ARAVQ används som datareducerare på en artificiell respektive en fysiologisk datamängd inom en datautvinningskontext. Dessa datamängder skiljer sig från tidigare robotikmiljöer i och med att de beskriver objekt med diffusa eller överlappande gränser i indatarymden. Varje datamängd klassificeras efter datareduceringen med hjälp av artificiella neuronnät. Resultatet från experimenten tyder på att klassificering med ARAVQ som datareducerare uppnår ett betydligt lägre resultat än om ARAVQ inte används som datareducerare. Detta antas delvis bero på den låga generaliserbarheten hos de lösningar som skapas av ARAVQ. I diskussionen föreslås att ARAVQ skall kompletteras med en funktion för grannskap, motsvarande den som finns i Self-Organizing Map. Med ett grannskap behålls relationerna mellan de kluster som ARAVQ skapar, vilket antas minska följderna av att en beskrivning hamnar i ett grannkluster

No description available.

Les récentes avancées technologiques des microphones de type microsystème électromécanique (MEMS) leurs permettent une utilisation sur une large gamme d’amplitudes sonores. Leur niveau de bruit ayant baissé, il devient possible de capter des sons provenant d’une distance plus lointaine, tandis que l’augmentation de leur pression acoustique maximale leur permet de ne pas saturer dans un environnement très bruyant de type concert ou évènement sportif. Ainsi le système électronique de conversion analogique-numérique connecté au microphone devient l’élément limitant les performances du système d’acquisition acoustique. Un besoin de nouvelles architectures de conversion analogique-numérique ayant une plage dynamique augmentée se fait donc ressentir. Par ailleurs, ces microphones étant de plus en plus utilisés dans des systèmes fonctionnant sur batterie, la contrainte de limitation de la consommation devient importante.Dans la bande de fréquences audio, les convertisseurs analogiques-numériques de type sigma-delta sont les plus aptes à obtenir une grande résolution combinée à une faible consommation. Ils sont divisés en deux grandes familles: ceux à temps discret utilisant principalement des circuits à capacités commutées, et ceux à temps continu utilisant des circuits classiques. Cette thèse se concentre sur l’étude et la conception de chacun des deux types de convertisseurs sigma delta, en insistant sur la faible consommation, le faible coût de production (surface occupée) et la robustesse du circuit, cela en vue d’une production de masse pour équipements portables.La conception d’un convertisseur analogique numérique de type sigma-delta à temps discret a été réalisé, ce dernier atteignant un rapport signal sur bruit de 100 décibels sur une bande de 24kHz, pour une puissance consommée de seulement 480μW. Pour limiter la consommation, de nouveaux amplificateurs à base d’inverseurs sont utilisés, et dont la robustesse contre les variations du procédé de fabrication ou de la température a été améliorée. Les spécifications ont été définies grâce au développement d’un modèle de haut-niveau précis, ce qui permet d’éviter le surdimensionnement tout en atteignant les performances voulues. Enfin, un grand ratio de suréchantillonnage a été choisi afin de réduire l’espace utilisé par les capacités commutées, minimisant le coût de fabrication.Après une étude théorique de l’équivalence entre les modulateurs sigma-delta à temps discret et à temps continu, ainsi que des spécificités propres aux modulateurs à temps continu, une réalisation de ces derniers a été effectuée. Celui-ci atteint un rapport signal sur bruit de 95 décibels sur une bande de fréquence de 24kHz, tout en consommant 142μW. Pour réduire la consommation ainsi que l’espace utilisé, un filtre de boucle du second-ordre a été réalisé avec un seul amplificateur, et le quantificateur fait aussi office d’intégrateur grâce à l’utilisation d’une structure d’oscillateurs contrôlés en tension. Ce quantificateur à base d’oscillateurs est réalisé par des cellules numériques, réduisant la consommation et l’espace utilisé, mais est hautement non-linéaire. Cette non-linéarité a été prise en compte par des choix architecturaux afin de ne pas réduire les performances finales du modulateur. / The recent technological advances in microelectromechanical system (MEMS) microphones allow them to be used on a large sound amplitude range. Due to their lower noise level, it becomes possible to capture sound from a faraway distance, while their increased acoustic overload point gives them the ability to capture sound without saturation in a loud environment like a concert or a sport event. Thus, the electronic analog / digital conversion system connected to the microphone becomes the limiting element of the acoustic acquisition system performance. There is then a need for a new analog / digital conversion architecture which has an increased dynamic range. Furthermore, since more and more of these microphones are used in battery-powered devices, the power consumption limitation constraint becomes of high importance.In the audio frequency band, the sigma-delta analog / digital converters are the ones most able to provide a high dynamic range combined to a limited power consumption. They are split in two families: the discrete-time ones using switched-capacitors circuits and the continuous-time ones using more classical structures. This thesis concentrates on the study and the design of both of these two types of sigma-delta converters, with an emphasis on the low-power consumption, the low production cost (area occupied) and the circuit robustness, in sight of a mass production for portable devices.A discrete-time sigma-delta modulator design has been made, the latter reaching a signal to noise ratio of 100dB on a 24kHz frequency bandwidth, for a power consumption of only 480μW. To limit the power consumption, new inverter-based amplifiers are used, with an improved robustness against the variations of the fabrication process or the temperature. Amplifier specifications are obtained thanks to an accurate high-level model developed, which allows to avoid over-design while ensuring that the wanted performances are reached. Finally, a large oversampling ratio has been used to reduce the switched-capacitors area, lowering the modulator cost.After a theoretical study of the equivalence between discrete-time and continuous-time modulators, and of continuous-time modulators specificities, a design of the latter has been made too. It reaches a signal to noise ratio of 95dB on a 24kHz bandwidth, while consuming 142μW. To reduce the power consumption and the occupied area, a second-order loop filter is implemented using a single amplifier, and the quantizer uses a VCO-based structure that provides inherently an integrating stage. The VCO-based quantizer is made using digital cells, lowering the consumption and area, but is highly non-linear. This non-linearity has been handled by architectural choices to not influence the final modulator performances.

This thesis was conducted in collaboration with Signal Processing (SP) Devices Sweden AB. SP Devices provides digital signal processing solutions for the enhancement of analogue to digital conversion (ADC). Their ADCs facilitate the development of products for Communications, Radio base stations, Radar, Signals intelligence and Test & Measurement. The ADQ series digitizers, from SP Devices, are portable high performance digitizers which incorporate one or more analog inputs, an on-board double data rate (DDR2) memory and USB or PXI Express interface. DDR2 refers to the ability of a computer bus to transfer data on both the rising and falling edges of a clock signal. The ADCaptureLab software is a graphical user interface used to control this digitizer. ADCaptureLab, designed in the C/C++ programming language, is an easy-to-use standalone program which allows for configuration and operation of all ADQ series digitizers from SP Devices. The use of the LabView program from National Instruments forms the backbone of this thesis. LabVIEW (short for Laboratory Virtual Instrumentation Engineering Workbench) is a platform and development environment for a visual programming language from National Instruments. The topic of this thesis was to reproduce the ADCaptureLab user interface using LabView instead of C/C++. The graphical user interface (GUI) developed in LabView should be able to communicate with and control the processing of the ADQ214 digitizer (the digitizer model provided to us) in the same way as the ADCaptureLab. This would involve not only the data capturing and visualization but also digitizer configurations, monitoring of the device and ADQ functions and the analyses of acquired signal and FFT. In order to implement the configuration settings we developed functions for trigger settings (conditions at which a trigger will occur), Analogue Front End settings (AC/DC coupling), clock settings (sets the clock source), data type settings (set sample format), gain and offset setting (sets amplitude gain and mean value), pre-trigger settings (size of pre-trigger buffer), trigger hold off settings ( number of samples to wait for acquiring data after trigger), data acquisition length settings (Length of the acquired signal), continuous and single batch data acquisition, FFT transformation, save, load , and “clear plots” control (resets graph indicators). The functioning of our device may be monitored through the “Status window” (displays connection status of the ADQ device), “Devices window” (displays product information about the ADQ device), “Device monitor window” (returns the status of ADQ-API functions used) and the “Device information window” (returns information related to the revision of the ADQ device). Analyzing the acquired data and its corresponding FFT is made simple with the “Signal Properties” window (displays analyzed data), “Mark Harmonics” control (marks harmonics in the FFT) and the “Mark Signal Props” control (marks the fundamental tone and highest distortion in the FFT). Our LabVIEW GUI efficiently incorporates the features described above. In addition to being able to communicate instructions to the ADQ214 device we are able to monitor its condition and analyze any output. This result serves to show that it is possible to develop a program such as ADCaptureLab in LabVIEW. / HASSAN GILANI # : +46736742637 RAISUL BHUIYAN # :+46762596979

H διδακτορική διατριβή μελετά ζητήματα που αφορούν την ενσωμάτωση υπολογιστικών μοντέλων ακοής για την μοντελοποίηση και επεξεργασία ηχητικών σηματών για την βέλτιστη αναπαραγωγή τους σε χώρους με αντήχηση καθώς και την κωδικοποίηση ηχητικών δεδομένων. Το κύριο μέρος της διατριβής επικεντρώθηκε στην μοντελοποίηση των αντιληπτικά σημαντικών αλλοιώσεων λόγω αντήχησης, με την βοήθεια κατάλληλα οριζόμενων μόνο-ωτικών και διαφορικών ενδο-καναλικών παραμέτρων και την απεικόνιση τους με τη βοήθεια χρονο-συχνοτικών 2Δ αναπαραστάσεων. Ο λεπτομερής εντοπισμός των αλλοιώσεων στα ηχητικά σήματα μέσω του προτεινόμενου Δείκτη Επικάλυψης λόγω Αντήχησης (ΔΕΑ) διαμόρφωσε κατάλληλη μεθοδολογία ανάλυσης-σύνθεσης, για την καταστολή της αντήχησης σε συγκεκριμένες χρονο-συχνοτικές περιοχές. Το κύριο πλεονέκτημα της προτεινόμενης, εξαρτώμενης του σήματος, μεθοδολογίας είναι ότι επιτυγχάνεται η καταστολή των, με σχετική καθυστέρηση, παραμορφώσεων λόγω αντήχησης σε μια μεγαλύτερη κλίμακα, δεδομένου ότι μόνο οι αντιληπτικά σημαντικές περιοχές του σήματος επηρεάζονται από την επεξεργασία. Επιπλέον, αναζητήθηκε η δυνατότητα ανάλυσης των ηχητικών δεδομένων με βάση τις εσωτερικές τους αναπαραστάσεις (όπως δηλαδή τις παρέχει το υπολογιστικό μοντέλο ακοής) με εφαρμογή στην περιοχή της κωδικοποίησης σημάτων. Ο προτεινόμενος μη-ομοιόμορφος κβαντιστής πραγματοποιεί τη διαδικασία της κβάντισης χρονο-συχνοτικά με κατάλληλη οδήγηση από το υπολογιστικό μοντέλο ακοής, εξασφαλίζοντας καλύτερη υποκειμενική ηχητική ποιότητα, σε σχέση με ένα ομοιόμορφο PCM κβαντιστή. Χρησιμοποιώντας τη βασική λειτουργία του μη-ομοιόμορφου κβαντιστή, υλοποιήθηκε ενά κριτήριο αξιολόγησης ηχητικών δεδομένων, όπου σε αντίθεση με καθιερώμενα κριτήρια (όπως το Noise to Mask Ration, NMR) επιτελεί τις λειτουργίες του στο πεδίο χρόνου-συχνότητας και παρέχει τη δυνατότητα εντοπισμού της υποκειμενικά σημαντικής παραμόρφωσης με βάση την χρονική εξέλιξη του σήματος. / The dissertation studies issues concerning the integration of computational auditory models for modeling and processing of audio signals for optimal reproduction in reverberant spaces as well as topics related to audio coding. Based on the theoretical framework analysis that was established, the necessity of a signal-dependent approach was underlined for modeling the perceptually-relevant effects of reverberation. The main part of the dissertation thesis was focused on describing the perceptually-relevant alterations due to reverberation, based on appropriate defined monaural and differential inter-channel parameters and also their representation with well-defined time-frequency 2D maps. The detailed localization of alterations due to reverberation in the acoustic signals via the proposed Reverberation Masking Index (RMI) introduced an analysis-synthesis methodology for the compensation of reverberation in perceptually-significant time-frequency regions incorporating also, well-established digital signal processing techniques. The main advantage of the proposed signal-dependent methodology is that the suppression of reverberant tails can be achieved on a larger scale under practical conditions, since only perceptually significant regions of the signal are affected after processing. Additionally, the proposed framework complements the more traditional system-dependent inverse filtering methods, enabling novel and efficient signal processing schemes to evolve for room dereverberation applications. The thesis examines also the feasibility of the acoustic signal analysis based on the internal representations provided by the computational auditory model, applicable in the area of audio coding. The proposed non-uniform quantizer operates in the time-frequency domain, where a novel quantization process is driven by the computational auditory model, thus enabling an overall better perceptual quality with respect to uniform PCM quantizer. Considering the fundamental operation of the novel non-uniform quantizer, a criterion for audio quality evaluation was proposed, where contrary to well-established criteria (i.e., Noise to Mask Ratio, NMR) its potential structure performs in the time-frequency domain and provides the detailed localization of perceptually-important distortions based on the input signal’s evolution.

<p>    </p> <p>This thesis studies networked Gaussian linear control systems with random delays. Networked control systems is a popular topic these years because of their versatile applications in daily life, such as smart grid and unmanned vehicles. With the development of these systems, researchers have explored this area in two directions. The first one is to derive the inherent rate-cost relationship in the systems, that is the minimal transmission rate needed to achieve an arbitrarily given stability requirement. The other one is to design achievability schemes, which aim at using as less as transmission rate to achieve an arbitrarily given stability requirement. In this thesis, we explore both directions. We assume the sensor-to-controller channels experience independently and identically distributed random delays of bounded support. Our work separates into two parts. In the first part, we consider networked systems with only one sensor. We focus on deriving a lower bound, R_{LB}(D), of the rate-cost tradeoff with the cost function to be E{| <strong>x^</strong>T<strong>x </strong>|} ≤ D, where <strong>x </strong>refers to the state to be controlled. We also propose an achievability scheme as an upper bound, R_{UB}(D), of the optimal rate-cost tradeoff. The scheme uses lattice quantization, entropy encoder, and certainty-equivalence controller. It achieves a good performance that roughly requires 2 bits per time slot more than R_{LB}(D) to achieve the same stability level. We also generalize the cost function to be of both the state and the control actions. For the joint state-and-control cost, we propose the minimal cost a system can achieve. The second part focuses on to the covariance-based fusion scheme design for systems with multiple > 1 sensors. We notice that in the multi-sensor scenario, the outdated arrivals at the controller, which many existing fusion schemes often discard, carry additional information. Therefore, we design an implementable fusion scheme (CQE) which is the MMSE estimator using both the freshest and outdated information at the controller. Our experiment demonstrates that CQE out-performances the MMSE estimator using the freshest information (LQE) exclusively by achieving a 15% smaller average L2 norm using the same transmission rate. As a benchmark, we also derive the minimal achievable L2 norm, Dmin, for the multi-sensor systems. The simulation shows that CQE approaches Dmin significantly better than LQE. </p>