Este trabalho apresenta a pesquisa sobre a extração de parâmetros característicos de sinais acústicos para fins de detecção automática de vazamento de água em tubulações enterradas. Os sinais acústicos foram adquiridos com o auxílio de um geofone eletrônico e também catalogados por técnicos especialistas em detecção acústica. De todos os sinais foram extraídos os modelos de predição linear perceptual de várias ordens, determinando-se como melhor a ordem 2. A partir de um conjunto de modelos de referência de sinais de vazamento, a distância média de Itakura dos outros modelos em relação a estas referências foram calculadas. Em conjunto com estas distâncias, quatro características espectrais são também extraídas do sinal a fim de compor o vetor de parâmetros característicos do sinal. Parte destes vetores de parâmetros característicos são utilizados para treinar o classificador de máquina de vetores de suporte. O restante dos dados são, então, submetidos a este classificador que obteve a taxa de acerto de classificação em torno de 93%. Experimentos anteriores, utilizando modelos de predição linear, de ordem 10, obtiveram uma taxa de acerto em torno de 82%. Isso demonstra que estes novos parâmetros característicos propostos alcançam os objetivos deste trabalho, que são algoritmos com melhor taxa de acerto na detecção de vazamentos. / This work presents a research about feature extraction of acoustic signals for detection of water leak in buried pipes. Acoustic signals were acquired by means of an electronic geophone and also labeled by technicians specialized in acoustic water leak detection. For every signals, its linear predictive model was estimated for a range of prediction orders, concluding for the best order 2. Out of this group of models, some leaky ones are used as reference for calculating the Itakura mean distance with respect to the other models. Completing this measure, four spectral features are extracted to compose the signal feature vector. Some of these vectors were used to train a support vector machine to be used as a classifier. The remaining ones were used to evaluate the classification. The resulting accuracy rate achieved is around 93%. Earlier experiments, which use linear prediction of order 10 had an accuracy rate around 82%. This shows that this novel proposal of feature vector achieves the main goal of this research, which is the increase in the leak detection accuracy rate.

Current high quality text-to-speech (TTS) systems are based on unit selection from a large database that is both contextually and prosodically rich. These systems, albeit capable of natural voice quality, are computationally expensive and require a very large footprint. Their success is attributed to the dramatic reduction of storage costs in recent times. However, for many TTS applications a smaller footprint is becoming a standard requirement. This thesis presents a new method for representing speech segments that can improve the quality and/or reduce the footprint current concatenative TTS systems. The circular linear prediction (CLP) model is revisited and combined with the constant pitch transform (CPT) to provide a robust representation of speech signals that allows for limited prosodic movements without a perceivable loss in quality. The CLP model assumes that each frame of voiced speech is an infinitely periodic signal. This assumption allows for LPC modeling using the covariance method, with the efficiency of the autocorrelation method. The CPT is combined with this model to provide a database that is uniform in pitch for matching the target prosody during synthesis. With this representation, limited prosody modifications and unit concatenation can be performed without causing audible artifacts. For resolving artifacts caused by pitch modifications in voicing transitions, a method has been introduced for reducing peakiness in the LP spectra by constraining the line spectral frequencies. Two experiments have been conducted to demonstrate the potential for the capabilities of CLP/CPT method. The first is a listening test to determine the ability of this model to realize prosody modifications without perceivable degradation. Utterances are resynthesized using the CLP/CPT method with emphasized prosodics to increase intelligibility in harsh environments. The second experiment compares the quality of utterances synthesized by unit-selection based limited-domain TTS against the CLP/CPT method. The results demonstrate that the CLP/CPT representation, applied to current concatenative TTS systems, can reduce the size of the database and increase the prosodic richness without noticeable degradation in voice quality.

We have devised a processor for pulsed Doppler radars for multi-target detection in same folded range under land and moving clutter. To this end, we have investigated the estimation of parameters, i.e., frequencies, amplitudes, and phases, of complex exponentials that model target echoes under radar clutter characterized by antenna scanning modulation with observation limited to single snapshot, i.e., one burst. The Maximum Likelihood method of estimation is presented together with the bounds on estimates, i.e., Cram&eacute / r-Rao bounds. We have analyzed linear prediction, together with its efficient implementation invented by Tufts &amp / Kumaresan, and compared its performance to other high resolution frequency estimation algorithms all modified to run under clutter. The essential part of the work is that line spectra estimation techniques model the clutter process also as a complex exponential. In addition, linear prediction combined with linear time&ndash / invariant maximum Signal to Interference Ratio (SIR) processor is analyzed. A technique to determine the model order, which is required by the frequency estimation algorithms, is presented that does not distinguish between targets and clutter. Clutter region concept is introduced to identify targets from clutter. The possibility to use these algorithms for target classification is briefly explained after providing a literature survey on helicopter echoes.

Η παρούσα διπλωματική εργασία έχει ως αντικείμενο την μελέτη και ανάπτυξη μίας τεχνικής ανίχνευσης φάσματος (spectrum sensing technique), η οποία να μπορεί να χρησιμοποιηθεί σε περιβάλλον Δυναμικής Εκχώρησης Φάσματος από Γνωστικά Συστήματα Ραδιοεπικοινωνιών (Cognitive Radio Systems). Οι παραδοσιακές στατικές στρατηγικές καταμερισμού του φάσματος έχουν δημιουργήσει προβλήματα έλλειψης διαθέσιμου φάσματος. Ταυτόχρονα, πρόσφατες μετρήσεις δείχνουν ότι μεγάλα τμήματα του φάσματος που έχουν εκχωρηθεί με άδεια σε συγκεκριμένα συστήματα υποχρησιμοποιούνται. Είναι επομένως αναγκαίο να υιοθετηθούν νέες πολιτικές διαχείρισης του φάσματος οι οποίες θα επιτρέπουν σε μη αδειοδοτημένα δίκτυα να κάνουν χρήση τμημάτων του αδειοδοτημένου φάσματος. Τα Γνωστικά Συστήματα Ραδιοεπικοινωνιών είναι ευφυή συστήματα τα οποία έχουν γνώση του περιβάλλοντός τους και μπορούν να προσαρμόζουν κατάλληλα τις παραμέτρους λειτουργίας τους σε αυτό. Τα συστήματα αυτά μπορούν να ανιχνεύουν περιοδικά το φάσμα, να εντοπίζουν τις ζώνες συχνοτήτων οι οποίες δε χρησιμοποιούνται από τους αδειοδοτημένους χρήστες τους και να τις αξιοποιούν. Όπως γίνεται εύκολα αντιληπτό από τα παραπάνω η ανίχνευση φάσματος αποτελεί ένα ιδιαιτέρως κρίσιμο θέμα για τα Γνωστικά Συστήματα Ραδιοεπικοινωνιών. Στο στάδιο αυτό, το σύστημα ανιχνεύει και παρακολουθεί στο περιβάλλον μέσα στο οποίο ενεργεί, το κατά πόσο το φάσμα είναι ελεύθερο ανά πάσα χρονική στιγμή και αξιοποιεί αυτά τα φασματικά κενά. Ουσιαστικά η ανίχνευση φάσματος εφαρμόζεται για να δώσει στον cognitive χρήστη μία όσο το δυνατόν πιστότερη εικόνα του περιβάλλοντος μέσα στο οποίο βρίσκεται. Η δική μας μελέτη επικεντρώθηκε στις τεχνικές ανίχνευσης φάσματος (spectrum sensing) και συγκεκριμένα αναπτύσσουμε μία μέθοδο ανίχνευσης φασματικών κενών βασιζόμενη στη χρήση ενός προβλεπτή (predictor) και στη χρησιμοποίηση του σφάλματος πρόβλεψης του σήματος που προκύπτει από αυτόν ως μετρική για τη λήψη απόφασης σχετικά με την ύπαρξη ή την απουσία σήματος ακόμα και σε θορυβώδη περιβάλλοντα (πολύ χαμηλό SNR). H τεχνική ανίχνευσης φάσματος που προτείνουμε μοντελοποιήθηκε στο περιβάλλον μοντελοποίησης MATLAB. Στη συνέχεια, διενεργήθηκαν εκτενείς προσομοιώσεις για ποικίλες τιμές των διαφόρων παραμέτρων του συστήματος αλλά και για διαφορετικά συστήματα, ούτως ώστε να αξιολογηθεί η επίδοση της τεχνικής σε διάφορες συνθήκες. / In the present thesis, we will study spectrum sensing techniques of Cognitive Radio SIMO systems. The conventional approach to spectrum management is not flexible, as most of the useful part of the spectrum is bounded. Hence it is extremely difficult to find free frequencies in order to deploy new services or to enhance the already existing ones. At the same time, various measurements show that the licensed spectrum is heavily underutilized in terms of both the time domain as well as the space domain. Thus Cognitive Radio technology comes to offer solutions, mainly with regard to the issues mentioned above, providing a dynamic utilization of the spectrum. Cognitive Radio has been proposed for lower priority secondary systems intending to improve spectral efficiency through spectrum sensing thus allowing these systems to transmit at frequency bands that are detected to be unused. As we can easily understand from the above, spectrum sensing is a critical issue for cognitive systems. In order to achieve adaptive transmission in unused portions of the spectrum without interferences to the licensed users of these portions (Primary Users-PUs), spectrum sensing is the first and one of the most important steps as high reliability is demanded on PUs' signal detection. That is, Secondary Users (SUs) should know if the spectrum is being used in order to exploit the available spectrum in the most efficient way. Essentially, spectrum sensing is used in order to provide the cognitive user with a representation of its operating environment which is as faithful as possible. The scope of this thesis is the study and the creation of algorithms that will give the SU of a SIMO system the opportunity to detect the existence of spectrum holes. The implementation we used is based on a predictor. More specifically, the received signal passes through a backward linear predictor from which we compute the difference between the actual signal and the predicted signal, which is the prediction error. By properly exploiting the prediction error, more precisely the power of the prediction error, we can trustworthily detect the existence or the absence of a signal, even in noisy environments, that is, for low values of the signal-to-noise ratio. In order to test the performance of our algorithms, the system above was simulated by MATLAB for different conditions and channels.

This diploma thesis deals with the theoretical analysis of the predictive methods of signal interpolation and signal modeling using sinusoidal model. On the basis of this theory the algorithm for the reconstruction of the missing sections in the audio signal is implemented in computing environment MATLAB. Results of mass testing reconstructions are displayed using objective methods SNR and PEMO-Q. Further experiments are carried out on single signals and their evaluation is described.

This thesis examines the benefits of using adaptive modulation and coding in terms of spectral efficiency and probability of bit error. Specifically, we examine the performance enhancement made possible by using linear prediction along with channel estimation in conjunction with adaptive modulation. We begin this manuscript with basic fundamentals of our study, followed by a detailed view of simulations, their results, and our conclusions from them. The study includes simulations in slow and moderately fast flat fading Rayleigh channels. We present our findings regarding the advantages of using predictive measures to foresee the state of the channel and make adjustments to transmissions accordingly. In addition to finding the general advantages of channel prediction in adaptive modulation, we explore various ways to adjust the prediction algorithm when we are faced with high Doppler rates and fast fading. By the end of this work, we should have a better understanding of when channel prediction is most valuable to adaptive modulation and when it is weakest, and how we can alleviate the problems that prediction will have in harsh environments. / Master of Science

Neste trabalho, foram utilizadas as vantagens da ferramenta matemática de análise temporal e espectral, a transformada wavelet discreta (DWT), além dos coeficientes de predição linear (LPC) e do algoritmo de inteligência artificial, Least Squares Support Vector Machines (LS-SVM), para aplicações em análise de sinais de voz e classificação de vozes patológicas. Inúmeros trabalhos na literatura têm demonstrado o grande interesse existente por ferramentas auxiliares ao diagnóstico de patologias da laringe. Os componentes da DWT forneceram parâmetros de medida para a análise e classificação das vozes patológicas, principalmente aquelas provenientes de pacientes com edema de Reinke e nódulo nas pregas vocais. O banco de dados com as vozes patológicas foi obtido do Departamento de Otorrinolaringologia e Cirurgia de Cabeça e Pescoço do Hospital das Clínicas da Faculdade de Medicina de Ribeirão Preto (FMRP-USP). Utilizando-se o algoritmo de reconhecimento de padrões, LS-SVM, mostrou-se que a combinação dos componentes da DWT de Daubechies com o filtro LP inverso levou a um classificador de bom desempenho alcançando mais de 90% de acerto na classificação das vozes patológicas. / The main objective of this work was to use the advantages of the time-frequency analysis mathematical tool, discrete wavelet transform (DWT), besides the linear prediction coefficients (LPC) and the artificial intelligence algorithm, Least Squares Support Vector Machines (LS-SVM), for applications in voice signal analysis and classification of pathological voices. A large number of works in the literature has been shown that there is a great interest for auxiliary tools to the diagnosis of laryngeal pathologies. DWT components gave measure parameters for the analysis and classification of pathological voices, mainly that ones from patients with Reinke\'s edema and nodule in the vocal folds. It was used a data bank with pathological voices from the Otolaryngology and the Head and Neck Surgery sector of the Clinical Hospital of the Faculty of Medicine at Ribeirão Preto, University of Sao Paulo (FMRP-USP), Brazil. Using the automatic learning algorithm applied in pattern recognition problems, LS-SVM, results have showed that the combination of Daubechies\' DWT components and inverse LP filter leads to a classifier with good performance reaching more than 90% of accuracy in the classification of the pathological voices.

In this thesis, we analyze the bandwidth requirements of MPEG-4 AVC video traffic and then propose and evaluate the accuracy of new MPEG-4 AVC video traffic models. First, we analyze the bandwidth requirements of the videos by comparing the statistical characteristics of the different frame types. We analyze their coefficient of variability, autocorrelation, and crosscorrelation in both short and long term. The Hurst parameter is also used to investigate the long range dependence of the video traces. We then provide an insight into B-frame dropping and its impact on the statistical characteristics of the video trace. This leads us to design two algorithms that predict the size of the B-frame and the size of the group of pictures (GOP) in the short-term. To evaluate the accuracy of the prediction, a model for the error is proposed. In a broadband cable network, B-frame size prediction can be employed by a cable headend to provision video bandwidth efficiently or more importantly, reduce bit rate variability and bandwidth requirements via selective B-frame dropping, thereby minimizing buffering requirements and packet losses at the set top box. It will be shown that the model provides highly accurate prediction, in particular for movies encoded in high quality resolution. The GOP size prediction can be used to provision bandwidth. We then enhance the B-frame and GOP size prediction models using a new scene change detector metric. Finally, we design an algorithm that predicts the size of different frame types in the long-term. Clearly, a long-term prediction algorithm may suffer degraded prediction accuracy and the higher complexity may result in higher latency. However, this is offset by the additional time available for long-term prediction and the need to forecast bandwidth usage well ahead of time in order to minimize packet losses during periods of peak bandwidth demands. We also analyze the impact of the video quality and the video standard on the accuracy of the model.

Neste trabalho, foram utilizadas as vantagens da ferramenta matemática de análise temporal e espectral, a transformada wavelet discreta (DWT), além dos coeficientes de predição linear (LPC) e do algoritmo de inteligência artificial, Least Squares Support Vector Machines (LS-SVM), para aplicações em análise de sinais de voz e classificação de vozes patológicas. Inúmeros trabalhos na literatura têm demonstrado o grande interesse existente por ferramentas auxiliares ao diagnóstico de patologias da laringe. Os componentes da DWT forneceram parâmetros de medida para a análise e classificação das vozes patológicas, principalmente aquelas provenientes de pacientes com edema de Reinke e nódulo nas pregas vocais. O banco de dados com as vozes patológicas foi obtido do Departamento de Otorrinolaringologia e Cirurgia de Cabeça e Pescoço do Hospital das Clínicas da Faculdade de Medicina de Ribeirão Preto (FMRP-USP). Utilizando-se o algoritmo de reconhecimento de padrões, LS-SVM, mostrou-se que a combinação dos componentes da DWT de Daubechies com o filtro LP inverso levou a um classificador de bom desempenho alcançando mais de 90% de acerto na classificação das vozes patológicas. / The main objective of this work was to use the advantages of the time-frequency analysis mathematical tool, discrete wavelet transform (DWT), besides the linear prediction coefficients (LPC) and the artificial intelligence algorithm, Least Squares Support Vector Machines (LS-SVM), for applications in voice signal analysis and classification of pathological voices. A large number of works in the literature has been shown that there is a great interest for auxiliary tools to the diagnosis of laryngeal pathologies. DWT components gave measure parameters for the analysis and classification of pathological voices, mainly that ones from patients with Reinke\'s edema and nodule in the vocal folds. It was used a data bank with pathological voices from the Otolaryngology and the Head and Neck Surgery sector of the Clinical Hospital of the Faculty of Medicine at Ribeirão Preto, University of Sao Paulo (FMRP-USP), Brazil. Using the automatic learning algorithm applied in pattern recognition problems, LS-SVM, results have showed that the combination of Daubechies\' DWT components and inverse LP filter leads to a classifier with good performance reaching more than 90% of accuracy in the classification of the pathological voices.

This thesis deals with description of human hearing, audio and speech codecs, description of objective measure of quality and practical comparison of codecs. Chapter about audio codecs consists of description of lossless codec FLAC and lossy codecs MP3 and Ogg Vorbis. In chapter about speech codecs is description of linear predictive coding and G.729 and OPUS codecs. Evaluation of quality consists of description of segmental signal-to- noise ratio and perceptual evaluation of quality – WSS and PESQ. Last chapter deals with description od practical part of this thesis, that is comparison of memory and time consumption of audio codecs and perceptual evaluation of speech codecs quality.