Power distribution network (PDN) design was a non-existent entity during the early days of microprocessors due to the low frequency of operation. Once the switching frequencies of the microprocessors started moving towards and beyond MHz regions, the parasitic inductance of the PCB tracks and planes started playing an important role in determining the maximum voltage on a PDN. Voltage regulator module (VRM) sup-plies only the DC power for microprocessors. When the MOSFETs inside a processor switches, it consumes currents during transition time. If this current is not provided, the voltage on the supply rails can go below the specifications of the processor. For lower MHz processors few ceramic-capacitors known as ‘decoupling capacitors’ were connected between power and ground to provide this transient current demand. When the processor frequency increased beyond MHz, the number of capacitors also increased from few numbers to hundreds of them. Nowadays, the PDN is said to be comprising all components from VRM till the die location. It includes VRM, bulk capacitors, PCB power planes, capacitor mounting pads and vias, mount for the electronic package, package capacitors, die mount and internal die capacitance. So, the PDN has evolved into a very complex system over the years. A PDN should provide three distinct roles; 1) provide transient current required by the processor 2) act as a stable reference voltage for processor 3) filter out the noise currents injected by the processor. The first two are required for the correct operation of the processor. Third one is a requirement from analog or other sensitive circuits connected to the same PDN. If the noise exits the printed circuit board (PCB), it can result in conducted and radiated EMI, which can in turn result in failure of a product in EMC testing. Every PDN design starts with the calculation of a target impedance which is given as the ratio of maximum allowed ripple voltage to the maximum transient current required by the processor. The transient current is usually taken as half the average input current. The definition of target impedance assumes that the PDN is flat over the entire frequency of operation, which is true only for a resistive network. This is seldom true for a practical PDN, since it contains inductances and capacitances. Because of this, a practical PDN has an uneven impedance versus frequency envelope. Whenever two capacitors with different self resonant frequencies are connected in parallel, their equivalent impedance produces a pole between the self resonant frequencies known as antiresonance peaks. Because of this, a PDN will have phase angles associated with them. Also, these antiresonance peaks are energy reservoirs which will be excited during the normal operation of a processor by the varying currents. The transient current of a microprocessor is modeled as a gamma function, but for practical cases it can be approximated as triangular waveforms during the transition time which is normally 10% of the time period. Depending upon the micro-operations running inside the processor, the peak value of this waveform varies. This is filtered by the on-chip capacitors, package inductance and package capacitors. Due to power gating, clock gating, IO operations, matrix multiplications and magnetic memory readings the waveforms at the board will be like pulse type, and their widths are determined by these operations. In literatures, these two types of waveforms are used for PDN analysis, depending upon at which point the study is conducted. Chapter 1 introduces the need for PDN design and the main roles of a PDN. The issue of antiresonance is introduced from a PDN perspective. Different types of capacitors used on a PDN are discussed with their strengths and limitations. The general nature of the switching noise injected by a microprocessor is also discussed. This chapter discusses the thesis contributions, and the existing work related to the field. Chapter 2 introduces a new method to calculate the target impedance (Zt ) by including the phase angles of a PDN which is based on a maximum voltage calculation. This new Zt equals to conventional Zt for symmetrical triangular switching current waveforms. The value of new Zt is less than the conventional Zt for trapezoidal excitation patterns. By adding the resonance effects into this, a maximum voltage value is obtained in this chapter. The new method includes the maximum voltage produced on a PDN when multiple antiresonance peaks are present. Example simulations are provided for triangular and pulse type excitations. A measured input current wave-form for PIC16F677 microcontroller driving eight IO ports is provided to prove the assumption of pulse type waveforms. For triangular excitation waveform, the maximum voltage predicted based on the expression was ¡0.6153 V, and the simulated maximum voltage was found to be at ¡0.5412 V which is less than the predicted value. But the predicted value based on Zt method was 1.9845 V. This shows that the conventional as well as the new target impedance method leads to over estimating the maximum voltage in certain cases. This is because most of the harmonics are falling on the minimum impedance values on a PDN. If the PDN envelope is changed by temperature and component tolerances, the maximum voltage can vary. So the best option is to design with the target impedance method. When pulse current excitation was studied for a particular PDN, the maximum voltage produced was -139.39 mV. The target impedance method produced a value of -100.24 mV. The maximum voltage predicted by the equation was -237 mV. So this shows that some times the conventional target impedance method leads to under estimating the PDN voltage. From the studies, it is shown that the time domain analysis is as important as frequency domain analysis. Another important observation is that the antiresonance peaks on a PDN should be damped both in number and peak value. Chapter 3 studies the antiresonance peak suppression methods for general cases. As discussed earlier, the antiresonance peaks are produced when two capacitors with different self resonant frequencies are connected in parallel. This chapter studies the effect of magnetic coupling between the mounting loops of two capacitors in parallel. The mounting loop area contribute to the parasitic inductance of a capacitor, and it is the major contributing factor to it. Other contributing factors are equivalent series inductance (ESL) and plane spreading inductance. The ESL depends on the size and on how the internal plates of the capacitors are formed. The spreading inductance is the inductance contributed by the parts of the planes connecting the capacitor connector vias to the die connections or to other capacitor vias. If the power and ground planes are closer, the spreading inductance is lower. On one/two layer boards dedicated power/ground planes are absent. So the spreading inductance is replaced by PCB track inductances. The inductance contributed by the mounted area of the capacitor is known as mounting inductance. On one/two layer boards dedicated power/ground planes are absent. So the spreading inductance is replaced by PCB track inductances. The dependencies of various circuit parameters on antiresonance peak are studied using circuit theory. A general condition for damping the antiresonance is formulated. The antiresonance peak reduces with Q factor. The conventional critical condition for antiresonance peak damping needs modification when magnetic coupling is present between the mounting loops of two parallel unequal value capacitors. By varying the connection geometry it is possible to obtain negative and positive coupling coefficients. The connection geometries to obtain these two are shown. An example is shown for positive and negative coupling coefficient cases with simulation and experimental results. For the example discussed, RC Æ 32 - for k Æ Å0.6 and RC Æ 64 - for k Æ ¡0.6, where RC is the critical damping value and k is the magnetic coupling coefficient between the two mounting loops. The reason for this is that, the antiresonance peak impedance value is higher for negative coupling coefficient case than that for positive coupling coefficient case. Above the self resonant frequencies of both the capacitors, the equivalent impedance of the parallel capacitors become inductive. This case is studied with two equal value capacitors in parallel. It is shown that the equivalent inductance is lower for negative coupling coefficient case as compared to positive coupling coefficient case. An example is provided with simulation and experimental results. In the experimental results, parasitic inductance is observed to be 2.6 times lower for negative coupling coefficient case than that for positive coupling coefficient case. When equal value capacitors are connected in parallel, it is advantageous to use a negative coupling geometry due to this. Chapter 4 introduces a new method to damp the antiresonance peak using a magnet-ically coupled resistive loop. Reducing the Q factor is an option to suppress the peak. In this new method, the Q factor reduction is achieved by introducing losses by mag-netically coupling a resistive loop. The proposed circuit is analyzed with circuit-theory, and governing equations are obtained. The optimum value of resistance for achieving maximum damping is obtained through analysis. Simulation and experimental results are shown to validate the theory. From the experimental results approximately 247 times reduction in antiresonance peak is observed with the proposed method. Effectiveness of the new method is limited by the magnetic coupling coefficient between the two mounting loops of capacitors. The method can be further improved if the coupling coefficient can be increased at the antiresonance frequency. Chapter 5 focuses on the third objective of a PDN, that is to reduce the noise injected by the microprocessor. A new method is proposed to reduce the conducted noise from a microprocessor with switched super capacitors. The conventional switched capacitor filters are based on the concept that the flying capacitor switching at high frequency looks like a resistor at low frequency. So for using at audio frequencies the flying capacitors were switching at MHz frequencies. In this chapter the opposite of this scenario is studied; the flying capacitors are the energy storage elements of a switched capacitor converter and they switch at lower frequencies as compared to the noise frequencies. Two basic circuits (1:1 voltage conversion ratio) providing noise isolation were discussed. They have distinct steady state input current waveforms and are explained with PSPICE simulations. The inrush current through switches are capable of destroying them in a practical implementation. A practical solution was proposed using PMOS-PNP pair. The self introduced switching noise of the converter is lower when switching frequency is low and turn ON-OFF time is higher. If power metal oxide semiconductor field effect transistor (MOSFET)s are used, the turn ON and turn OFF are slow. The switching frequency can be lowered based on the voltage drop power loss. The governing equations were formulated and simulated. It is found that the switching frequency can be lowered by increasing the capacitance value without affecting the voltage drop and power loss. From the equations, it is found that the design parameters have a cyclic dependency. Noise can short through the parasitic capacitance of the switches. Two circuits were proposed to improve the noise isolation: 1) T switch 2) ¦ switch. Of these, the ¦ switch has the higher measured transfer impedance. Experimental results showed a noise reduction of (40-20) dB for the conducted frequency range of 150 kHz - 30 MHz with the proposed 1:1 switched capacitor converter. One possible improvement of this method is to combine the noise isolation with an existing switched capacitor converter (SCC) topology. The discussed example had a switching frequency of 700 Hz, and it is shown that this can isolate the switching noise in kHz and MHz regions. In a PDN there are antiresonance peaks in kHz regions. If the proposed circuit is kept close to a microprocessor, it can reduce the excitation currents of these low frequency antiresonance peaks. Chapter 6 concludes the thesis by stating the major contributions and applications of the concepts introduced in the thesis. This chapter also discusses the future scope of these concepts.

This thesis deals with inertial navigation systems issues. It describes basics of reference frames, coordinate systems and matrix calculations for AHRS. There are also basic information about inertial sensors, inertial measurements units and its mistakes. One of the purposes of this paper could be explanation of inertial navigation systems terms. The main object of this thesis is to explore the influence of using multiple sensors of same type to enhance measurements of AHRS systems.

This diploma thesis deals with methods of noise suppression for speech recognition systems. In theoretical part are discussed basic terms of this topic and also methods for noise suppression. These are spectral subtraction, Wiener filtering, RASTA, mapping of spectrogram or algorithms based on noise estimation. In second part types of noise are analyzed, there is proposal and implementation of spectral subtraction method of noise suppression for speech recognition system. Also extensive testing of spectral subtractive algorithms in comparison with Wiener filter is conducted. Assessment of this testing is done with defined metrics, successfulness of recognition, recognition system score and signal to noise ratio.

In x-ray imaging, the x-ray radiation must be the minimum necessary to achieve the required diagnostic objective, to ensure the patients safety. However, low-dose acquisitions yield images with low quality, which affect the radiologists image interpretation. Therefore, there is a compromise between image quality and radiation dose. This work proposes an image restoration framework capable of restoring low-dose acquisitions to achieve the quality of full-dose acquisitions. The contribution of the new method includes the capability of restoring images with quantum and electronic noise, pixel offset and variable detector gain. To validate the image processing chain, a simulation algorithm was proposed. The simulation generates low-dose DBT projections, starting from fulldose images. To investigate the feasibility of reducing the radiation dose in breast cancer screening programs, a simulated pre-clinical trial was conducted using the simulation and the image processing pipeline proposed in this work. Digital breast tomosynthesis (DBT) images from 72 patients were selected, and 5 human observers were invited for the experiment. The results suggested that a reduction of up to 30% in radiation dose could not be perceived by the human reader after the proposed image processing pipeline was applied. Thus, the image processing algorithm has the potential to decrease radiation levels in DBT, also decreasing the cancer induction risks associated with the exam. / Em programas de rastreamento de câncer de mama, a dose de radiação deve ser mantida o mínimo necessário para se alcançar o diagnóstico, para garantir a segurança dos pacientes. Entretanto, imagens adquiridas com dose de radiação reduzida possuem qualidade inferior. Assim, existe um equilíbrio entre a dose de radiação e a qualidade da imagem. Este trabalho propõe um algoritmo de restauração de imagens capaz de recuperar a qualidade das imagens de tomossíntese digital mamária, adquiridas com doses reduzidas de radiação, para alcançar a qualidade de imagens adquiridas com a dose de referência. As contribuições do trabalho incluem a melhoria do modelo de ruído, e a inclusão das características do detector, como o ganho variável do ruído quântico. Para a validação a cadeia de processamento, um método de simulação de redução de dose de radiação foi proposto. Para investigar a possibilidade de redução de dose de radiação utilizada na tomossíntese, um estudo pré-clínico foi conduzido utilizando o método de simulação proposto e a cadeia de processamento. Imagens clínicas de tomossíntese mamária de 72 pacientes foram selecionadas e cinco observadores foram convidados para participar do estudo. Os resultados sugeriram que, após a utilização do processamento proposto, uma redução de 30% de dose de radiação pôde ser alcançada sem que os observadores percebessem diferença nos níveis de ruído e borramento. Assim, o algoritmo de processamento tem o potencial de reduzir os níveis de radiação na tomossíntese mamária, reduzindo também os riscos de indução do câncer de mama.

Η παρούσα διατριβή πραγματεύεται την αναγνώριση ομιλητή σε πραγματικές συνθήκες. Τα κύρια σημεία της εργασίας είναι: (1) αξιολόγηση διαφόρων προσεγγίσεων εξαγωγής χαρακτηριστικών παραμέτρων ομιλίας, (2) μείωση της ισχύος της περιβαλλοντικής επίδρασης στην απόδοση της αναγνώρισης ομιλητή, και (3) μελέτη τεχνικών κατηγοριοποίησης, εναλλακτικών προς τις υπάρχουσες. Συγκεκριμένα, στο (1), προτείνεται μια νέα δομή εξαγωγής παραμέτρων ομιλίας βασισμένη σε πακέτα κυματομορφών, κατάλληλα σχεδιασμένη για αναγνώριση ομιλητή. Εξάγεται με ένα αντικειμενικό τρόπο σε σχέση με την απόδοση αναγνώρισης ομιλητή, σε αντίθεση με την MFCC προσέγγιση, που βασίζεται στην προσέγγιση της αντίληψης της ανθρώπινης ακοής. Έπειτα, στο (2), δίνεται μια δομή για την εξαγωγή παραμέτρων βασισμένη στα MFCC, ανεκτική στο θόρυβο, για την βελτίωση της απόδοσης της αναγνώρισης ομιλητή σε πραγματικό περιβάλλον. Συνοπτικά, μια τεχνική μείωσης του θορύβου βασισμένη σε μοντέλο προσαρμοσμένη στο πρόβλημα της επιβεβαίωσης ομιλητή ενσωματώνεται απευθείας στη δομή υπολογισμού των MFCC. Αυτή η προσέγγιση επέδειξε σημαντικό πλεονέκτημα σε πραγματικό και ταχέως μεταβαλλόμενο περιβάλλον. Τέλος, στο (3), εισάγονται δύο νέοι κατηγοριοποιητές που αναφέρονται ως Locally Recurrent Probabilistic Neural Network (LR PNN), και Generalized Locally Recurrent Probabilistic Neural Network (GLR PNN). Είναι υβρίδια μεταξύ των Recurrent Neural Network (RNN) και Probabilistic Neural Network (PNN) και συνδυάζουν τα πλεονεκτήματα των γεννετικών και διαφορικών προσσεγγίσεων κατηγοριοποίησης. Επιπλέον, τα νέα αυτά νευρωνικά δίκτυα είναι ευαίσθητα σε παροδικές και ειδικές συσχετίσεις μεταξύ διαδοχικών εισόδων, και έτσι, είναι κατάλληλα για να αξιοποιήσουν την συσχέτιση παραμέτρων ομιλίας μεταξύ πλαισίων ομιλίας. Κατά την εξαγωγή των πειραμάτων, διαφάνηκε ότι οι αρχιτεκτονικές LR PNN και GLR PNN παρέχουν καλύτερη απόδοση, σε σχέση με τα αυθεντικά PNN. / This dissertation dials with speaker recognition in real-world conditions. The main accent falls on: (1) evaluation of various speech feature extraction approaches, (2) reduction of the impact of environmental interferences on the speaker recognition performance, and (3) studying alternative to the present state-of-the-art classification techniques. Specifically, within (1), a novel wavelet packet-based speech features extraction scheme fine-tuned for speaker recognition is proposed. It is derived in an objective manner with respect to the speaker recognition performance, in contrast to the state-of-the-art MFCC scheme, which is based on approximation of human auditory perception. Next, within (2), an advanced noise-robust feature extraction scheme based on MFCC is offered for improving the speaker recognition performance in real-world environments. In brief, a model-based noise reduction technique adapted for the specifics of the speaker verification task is incorporated directly into the MFCC computation scheme. This approach demonstrated significant advantage in real-world fast-varying environments. Finally, within (3), two novel classifiers referred to as Locally Recurrent Probabilistic Neural Network (LR PNN), and Generalized Locally Recurrent Probabilistic Neural Network (GLR PNN) are introduced. They are hybrids between Recurrent Neural Network (RNN) and Probabilistic Neural Network (PNN) and combine the virtues of the generative and discriminative classification approaches. Moreover, these novel neural networks are sensitive to temporal and special correlations among consecutive inputs, and therefore, are capable to exploit the inter-frame correlations among speech features derived for successive speech frames. In the experimentations, it was demonstrated that the LR PNN and GLR PNN architectures provide benefit in terms of performance, when compared to the original PNN.

This thesis presents novel signal processing algorithms for speech and music dereverberation. The proposed algorithms focus on blind single-channel suppression of late reverberation; however binaural and semi-blind methods have also been introduced. Late reverberation is a particularly harmful distortion, since it significantly decreases the perceived quality of the reverberant signals but also degrades the performance of Automatic Speech Recognition (ASR) systems and other speech and music processing algorithms. Hence, the proposed deverberation methods can be either used as standalone enhancing techniques or implemented as preprocessing schemes prior to ASR or other applied systems. The main dereverberation method proposed here is a blind dereverberation technique based on perceptual reverberation modeling has been developed. This technique employs a computational auditory masking model and locates the signal regions where late reverberation is audible, i.e. where it is unmasked from the clean signal components. Following a selective signal processing approach, only such signal regions are further processed through sub-band gain filtering. The above technique has been evaluated for both speech and music signals and for a wide range of reverberation conditions. In all cases it was found to minimize the processing artifacts and to produce perceptually superior clean signal estimations than any other tested technique. Moreover, extensive ASR tests have shown that it significantly improves the recognition performance, especially in highly reverberant environments. / Η διατριβή αποτελείται από εννιά κεφάλαια, δύο παραρτήματα καθώς και την σχετική βιβλιογραφία. Είναι γραμμένη στα αγγλικά ενώ περιλαμβάνει και ελληνική περίληψη. Στην παρούσα διατριβή, αναπτύσσονται μεθόδοι ψηφιακής επεξεργασίας σήματος για την αφαίρεση αντήχησης από σήματα ομιλίας και μουσικής. Οι προτεινόμενοι αλγόριθμοι καλύπτουν ένα μεγάλο εύρος εφαρμογών αρχικά εστιάζοντας στην τυφλή (“blind”) αφαίρεση για μονοκαναλικά σήματα. Στοχεύοντας σε πιο ειδικά σενάρια χρήσης προτείνονται επίσης αμφιωτικοί αλγόριθμοι αλλά και τεχνικές που προϋποθέτουν την πραγματοποίηση κάποιας ακουστικής μέτρησης. Οι αλγόριθμοι επικεντρώνουν στην αφαίρεση της καθυστερημένης αντήχησης που είναι ιδιαίτερα επιβλαβής για την ποιότητα σημάτων ομιλίας και μουσικής και μειώνει την καταληπτότητα της ομιλίας. Επίσης, επειδή αλλοιώνει σημαντικά τα στατιστικά των σημάτων, μειώνει σημαντικά την απόδοση συστημάτων αυτόματης αναγνώρισης ομιλίας καθώς και άλλων αλγορίθμων ψηφιακής επεξεργασίας ομιλίας και μουσικής. Έτσι οι προτεινόμενοι αλγόριθμοι μπορούν είτε να χρησιμοποιηθούν σαν αυτόνομες τεχνικές βελτίωσης της ποιότητας των ακουστικών σημάτων είτε να ενσωματωθούν σαν στάδια προ-επεξεργασίας σε άλλες εφαρμογές. Η κύρια μέθοδος αφαίρεσης αντήχησης που προτείνεται στην διατριβή, είναι βασισμένη στην αντιληπτική μοντελοποίηση και χρησιμοποιεί ένα σύγχρονο ψυχοακουστικό μοντέλο. Με βάση αυτό το μοντέλο γίνεται μία εκτίμηση των σημείων του σήματος που η αντήχηση είναι ακουστή δηλαδή που δεν επικαλύπτεται από το ισχυρότερο σε ένταση καθαρό από αντήχηση σήμα. Η συγκεκριμένη εκτίμηση οδηγεί σε μία επιλεκτική επεξεργασία σήματος όπου η αφαίρεση πραγματοποιείται σε αυτά και μόνο τα σημεία, μέσω πρωτότυπων υβριδικών συναρτήσεων κέρδους που βασίζονται σε δείκτες αντικειμενικής και υποκειμενικής αλλοίωσης. Εκτεταμένα αντικειμενικά και υποκειμενικά πειράματα δείχνουν ότι η προτεινόμενη τεχνική δίνει βέλτιστες ποιοτικά ανηχωικές εκτιμήσεις ανεξάρτητα από το μέγεθος του χώρου.

H παρούσα διατριβή εξετάζει το πρόβλημα της διαρροής μικροφώνου, δηλαδή την αλληλεπίδραση και παρεμβολή μεταξύ ταυτόχρονα ενεργών ηχητικών πηγών σε πολυκαναλικές ηχητικές διατάξεις. Παρ' όλο που είναι ένα πολύ συχνό φαινόμενο με το οποίο οι μηχανικοί ήχου έρχονται αντιμέτωποι καθημερινά, δεν έχουν προταθεί μέθοδοι επεξεργασίας σήματος για την επίλυση του προβλήματος. Εδώ, το πρόβλημα διατυπώνεται για πρώτη φορά στο πλαίσιο της επεξεργασίας σήματος. Αρχικά, διατυπώνεται στο πλαίσιο του τυφλού διαχωρισμού πηγών (blind source separation) και αναλύονται οι περιορισμοί αυτής της προσέγγισης. Στην συνέχεια, το πρόβλημα επαναδιατυπώνεται σαν πρόβλημα σήματος υπό θόρυβο στα πλαίσια της καταστολής θορύβου. Ένα πρωτότυπο γενικευμένο πλαίσιο καταστολής διαρροής μικροφώνου εξάγεται βασιζόμενο σε ένα φίλτρο Wiener με πολυκαναλικό όρο θορύβο, καθώς και την ευρέως χρησιμοποιούμενη τεχνική «κοντινού μικροφώνου». Το ακουστικό σύστημα που μοντελοποιεί την διαδικασία μίξης και αλληλεπίδρασης των πηγών αναλύεται και γίνεται διαχωρισμός των σχετικών κρουστικών αποκρίσεων χώρου (room impulse responses) σε απ' ευθείας ακουστικά μονοπάτια και ακουστικά μονοπάτια διαρροής. Οι ιδιότητες του απ' ευθείας ακουστικού μονο- πατιού, δηλαδή της απόκρισης «κοντινού μικροφώνου» αναλύονται για πρώτη φορά από την προσέγγιση της επεξεργασίας σήματος και της ακουστικής κλειστών χώρων για πρώτη φορά. Οι ιδιότητες του ακουστικού μονοπατιού διαρροής αναλύονται επίσης για πρώτη φορά με την χρήση ακουστικών παραμέτρων. Έχοντας καθορίσει τις βασικές ιδιότητες του ακουστικού συστήματος, μια μέθοδος για την καταστολή διαρροής μικροφώνου αναπτύσσεται για μια διάταξη δύο καναλιών, βασισμένη σε ένα φίλτρο Wiener και μια άμεση εκτίμηση των σχετικών πυκνοτήτων φασματικής ενέργεiας (power spectral density). Η απόδοση της μεθόδου για ηχογραφήσεις σε πραγματικούς χώρους είναι πολύ ικανοποιητική και με βάση αυτά τα αποτελέσματα, η μέθοδος επεκτείνεται για περισσότερες από δύο πηγές και μικρόφωνα σε αυθαίρετες διατάξεις. Η ολοκληρωμένη μέθοδος είναι τυφλή και αυτόματη, καθώς δεν απαιτεί την επέμβαση του χρήση. Δεν κάνει χρήση πρότερης γνώσης ούτε απαιτεί εκπαίδευση και είναι υπολογιστικά απλή. Προτείνεται επίσης μια πρωτότυπη μέθοδος ανίχνευσης χρονικών διαστημάτων όπου μόνο μια πηγή είναι ενεργή (χρονικά διαστήματα «σόλο»), η οποία επιτρέπει την εκτίμηση συντελεστών στάθμισης οι οποίοι αντιστοιχούν στην σχετική μείωση της ηχητικής στάθμης που υφίσταται κάθε ηχητική πηγή καθώς το σήμα διαδίδεται προς τα μικρόφωνα. Αυτή η μέθοδος σε συνδυασμό με μια νεά, πρωτότυπη τεχνική εκτίμησης των πυκνοτήτων φασματικής ενέργεαις, η οποία βασίζεται στην αναγνώριση των κυρίαρχων διακριτών συχνοτήτων, επιτρέπει την εκτίμηση όλων των σχετικών ποσοτήτων σε μια πολυκαναλική ηχητική διάταξη. Από αυτές υπολογίζεται ένα πολυκαναλικό φίλτρο Wiener για κάθε σήμα μικροφώνου, το οποίο δίνει την εκτίμηση του αντίστοιχου σήματος πηγής. / This thesis examines the problem of microphone leakage, that is the interference between simultaneously active sound sources in multichannel audio applications. Despite being a common problem with which sound engineers are confronted every day, almost no signal processing methods have been proposed to address this issue. In this work, the problem is formulated for the first time in a signal processing framework. First, it formulated inside the blind source separation (BSS) context and the limitations of related methods are analysed and reported. Since, BSS methods seem to be inappropriate for this specific problem, it is reformulated as a signal in noise problem inside the well-known noise suppression framework. Based on the widely adopted close-microphone technique a novel, generalized framework for leakage suppression is derived based on a multichannel Wiener filter. The acoustic system that models the mixing process is analysed and the related room impulse responses are discerned in direct and leakage acoustic paths. The properties of the direct acoustic path, that is the close-microphone response are investigated for the first time, from a signal processing point of view as well as a room acoustics perspective. The properties of the leakage acoustic path are also analysed for the first time using room acoustic parameters. After key properties of the acoustic paths have been identified, a method for the suppression of microphone leakage in a two channel audio setup is developed based on aWiener filter and a crude approximation of the related power spectral densities (PSDs). The performance of this method for actual recordings in real reverberant environments is more than adequate and based on these results, the method is extended for more than two sources and microphones in arbitrary arrangements. The complete method is blind and automatic, since it does not require any user input. It does not assume any prior knowledge or require training and is computationally efficient. A novel solo detection method has been developed that allows the estimation of weighting coefficients that correspond to the relative attenuation experienced by sound sources as they travel to each microphone. Combined with a new and advanced PSD estimation method based on the identification of dominant frequency bins, the related PSDs in a multichannel audio application can be identified. From these an appropriate multichannel Wiener filter for each microphone signal can be calculated, which will provide the estimated source signal at its output.

This dissertation presents algorithms for restoring some of the major corruptions observed in archived film or video material. The two principal problems of impulsive distortion (Dirt and Sparkle or Blotches) and noise degradation are considered. There is also an algorithm for suppressing the inter-line jitter common in images decoded from noisy video signals. In the case of noise reduction and Blotch removal the thesis considers image sequences to be three dimensional signals involving evolution of features in time and space. This is necessary if any process presented is to show an improvement over standard two-dimensional techniques. It is important to recognize that consideration of image sequences must involve an appreciation of the problems incurred by the motion of objects in the scene. The most obvious implication is that due to motion, useful three dimensional processing does not necessarily proceed in a direction 'orthogonal' to the image frames. Therefore, attention is given to discussing motion estimation as it is used for image sequence processing. Some discussion is given to image sequence models and the 3D Autoregressive model is investigated. A multiresolution BM scheme is used for motion estimation throughout the major part of the thesis. Impulsive noise removal in image processing has been traditionally achieved by the use of median filter structures. A new three dimensional multilevel median structure is presented in this work with the additional use of a detector which limits the distortion caused by the filters . This technique is found to be extremely effective in practice and is an alternative to the traditional global median operation. The new median filter is shown to be superior to those previously presented with respect to the ability to reject the kind of distortion found in practice. A model based technique using the 3D AR model is also developed for detecting and removing Blotches. This technique achieves better fidelity at the expense of heavier computational load. Motion compensated 3D IIR and FIR Wiener filters are investigated with respect to their ability to reject noise in an image sequence. They are compared to several algorithms previously presented which are purely temporal in nature. The filters presented are found to be effective and compare favourably to the other algorithms. The 3D filtering process is superior to the purely temporal process as expected. The algorithm that is presented for suppressing inter-line jitter uses a 2D AR model to estimate and correct the relative displacements between the lines. The output image is much more satisfactory to the observer although in a severe case some drift of image features is to be expected. A suggestion for removing this drift is presented in the conclusions. There are several remaining problems in moving video. In particular, line scratches and picture shake/roll. Line scratches cannot be detected successfully by the detectors presented and so cannot be removed efficiently. Suppressing shake and roll involves compensating the entire frame for motion and there is a need to separate global from local motion. These difficulties provide ample opportunity for further research.

In x-ray imaging, the x-ray radiation must be the minimum necessary to achieve the required diagnostic objective, to ensure the patients safety. However, low-dose acquisitions yield images with low quality, which affect the radiologists image interpretation. Therefore, there is a compromise between image quality and radiation dose. This work proposes an image restoration framework capable of restoring low-dose acquisitions to achieve the quality of full-dose acquisitions. The contribution of the new method includes the capability of restoring images with quantum and electronic noise, pixel offset and variable detector gain. To validate the image processing chain, a simulation algorithm was proposed. The simulation generates low-dose DBT projections, starting from fulldose images. To investigate the feasibility of reducing the radiation dose in breast cancer screening programs, a simulated pre-clinical trial was conducted using the simulation and the image processing pipeline proposed in this work. Digital breast tomosynthesis (DBT) images from 72 patients were selected, and 5 human observers were invited for the experiment. The results suggested that a reduction of up to 30% in radiation dose could not be perceived by the human reader after the proposed image processing pipeline was applied. Thus, the image processing algorithm has the potential to decrease radiation levels in DBT, also decreasing the cancer induction risks associated with the exam. / Em programas de rastreamento de câncer de mama, a dose de radiação deve ser mantida o mínimo necessário para se alcançar o diagnóstico, para garantir a segurança dos pacientes. Entretanto, imagens adquiridas com dose de radiação reduzida possuem qualidade inferior. Assim, existe um equilíbrio entre a dose de radiação e a qualidade da imagem. Este trabalho propõe um algoritmo de restauração de imagens capaz de recuperar a qualidade das imagens de tomossíntese digital mamária, adquiridas com doses reduzidas de radiação, para alcançar a qualidade de imagens adquiridas com a dose de referência. As contribuições do trabalho incluem a melhoria do modelo de ruído, e a inclusão das características do detector, como o ganho variável do ruído quântico. Para a validação a cadeia de processamento, um método de simulação de redução de dose de radiação foi proposto. Para investigar a possibilidade de redução de dose de radiação utilizada na tomossíntese, um estudo pré-clínico foi conduzido utilizando o método de simulação proposto e a cadeia de processamento. Imagens clínicas de tomossíntese mamária de 72 pacientes foram selecionadas e cinco observadores foram convidados para participar do estudo. Os resultados sugeriram que, após a utilização do processamento proposto, uma redução de 30% de dose de radiação pôde ser alcançada sem que os observadores percebessem diferença nos níveis de ruído e borramento. Assim, o algoritmo de processamento tem o potencial de reduzir os níveis de radiação na tomossíntese mamária, reduzindo também os riscos de indução do câncer de mama.

Tato diplomová práce se zabývá problematikou implementace Hands-Free komunikačních systémů v automobilovém průmyslu. První kapitola je zaměřena na teoretické aspekty zpracování řeči v embedded aplikacích, jako je potlačení šumu, potlačení akustické zpětné vazby a další faktory ovlivňující kvalitu Hands-Free systémů. Druhá kapitola obsahuje návrh kompaktního flexibilního mobilního testovacího zařízení pro bezdrátové komunikační Hands-Free moduly.