No description available.

This thesis describes the development of several small signal analysis methods for the modelling of power electronic converters. The methods are written generally and are intended to be able to be applied to all converter classes. In the penultimate chapter these general models are used to model the capacitor commutated converter. All the contained methods are based around a time domain small signal model. This time domain small model is a linearization of a power electronic system of passive components and ideal switches described as a hybrid system. The key problem in the derivation of the small signal model is the correct determination and description of the linearized effect of switching instant variation. Three analysis methods based upon the small signal model are advanced in this thesis, these are; time domain sensitivity matrices for use in a Newton determination of the cyclic steady state of a power electronic converter, partial waveform construction of harmonic sensitivity matrices for studying sensitivity of converters to harmonic disturbances, and harmonic state space models also for the construction of harmonic sensitivity matrices and for study of dynamic systems. Each modelling technique is applied to the more common converter topologies of the Buck-Boost converter and the Graetz Bridge before being finally applied to the capacitor commutated converter. Each technique is compared to PSCAD-EMTDC simulations for verification.

Separating unknown signal mixtures into their constituent parts is a difficult problem in signal processing called blind source separation. One of the benchmark problems in this area is the extraction of the fetal heartbeat from an electrocardiogram in which it is overshadowed by a strong maternal heartbeat. This thesis presents a study of a signal separation technique called independent component analysis (ICA), in order to assess its suitability for the maternal-fetal ECG separation problem. This includes an analysis of ICA on deterministic, stochastic, simulated and recorded ECG signals. The experiments presented in this thesis demonstrate that ICA is effective on linear mixtures of known simulated or recorded ECGs. The performance of ICA was measured using visual comparison, heart rate extraction, and energy, information theoretic, and fractal-based measures. ICA extraction of clinically recorded maternal-fetal ECGs mixtures, in which the source signals were unknown, were successful at recovering the fetal heart rate.

No description available.

No description available.

Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2006. / Speech is the primary and most natural means of communication between human beings. With the rapid spread of technology across the globe and the increased number of personal and public applications for digital equipment in recent years, the need for human/machine interaction has increased dramatically. Synthetic speech is audible speech produced by a machine automatically. A text-to-speech (TTS) system is one that converts bodies of text into digital speech signals which can be heard and understood by a person. Current TTS systems generally require large annotated speech corpora in the languages for which they are developed. For many languages these resources are not available. In their absence, a TTS system generates synthetic speech by means of mathematical algorithms constrained by certain rules. This thesis describes the design and implementation of a rule-based speech generation algorithm for use in a TTS system. The system allows the type, emphasis, pitch and other parameters associated with a sound and its particular mode of articulation to be specified. However, no attempt is made to model prosodic and other higher-level information. Instead, this is assumed known. The algorithm uses linear predictive (LP) models of monophone speech units, which greatly reduces the amount of data required for development in a new language. A novel approach to the interpolation of monophone speech units is presented to allow realistic transitions between monophone units. Additionally, novel algorithms for estimation and modelling of the harmonic and stochastic content of an excitation signal are presented. This is used to determine the amount of voiced and unvoiced energy present in individual speech sounds. Promising results were obtained when evaluating the developed system’s South African English speech output using two widely used speech intelligibility tests, namely the modified rhyme test (MRT) and semantically unpredictable sentences (SUS).

The performance of III-V heterojunction bipolar transistors (HBTs) has improved significantly over the past two decades. Today’s state of the art Indium Phosphide (InP) HBTs have a maximum frequency of oscillation greater than 800 GHz and have been used to realize an amplifier operating above 600 GHz . In comparison to silicon (Si) based devices, III-V HBTs have superior transport properties that enables a higher gain, higher speed, and noise performance, and much higher Johnson figure- of-merit . From this perspective, the InP HBT is one of the most promising candidates for high performance mixed signal electronic systems.

Ce travail entreprend d'évaluer l'évolution de l'acquisition phonologique par des étudiants français des contrastes anglais /ɪ/-/i:/ et /ʊ/-/u:/. Le corpus étudié provient d'enregistrements de conversations spontanées menées avec des étudiants natifs. 12 étudiants, 9 femmes et 3 hommes,ont été suivis lors de 4 sessions espacées chacune d'un intervalle de six mois. L'approche adoptée est résolument quantitative, et agnostique quant aux théories d'acquisition d'une deuxième langue (par exemple Flege 2005, Best 1995,Kuhl 2008). Afin d'estimer les éventuels changements de prononciation, une procédure automatique d'alignement et d'extraction des données acoustiques a été conçue à partir du logiciel PRAAT (Boersma 2001). Dans un premier temps, deux autres logiciels (SPPAS et P2FA, Bigi 2012 et Yuan &Liberman 2008) avaient aligné les transcriptions des enregistrements au phonème près. Plus de 90 000 voyelles ont ainsi été analysées. Les données extraites sont constituées d'informations telles que le nombre de syllabes du mot, de sa transcription acoustique dans le dictionnaire, de la structure syllabique, des phonèmes suivant et précédant la voyelle, de leur lieu et manière d'articulation, de leur appartenance ou non au même mot, mais surtout des relevés formantiques de F0, F1, F2, F3 et F4. Ces relevés formantiques ont été effectués à chaque pourcentage de la durée de la voyelle afin de pouvoir tenir compte des influences des environnements consonantiques sur ces formants. Par ailleurs, des théories telles que le changement spectral inhérent aux voyelles (Nearey & Assmann(1986), Morrison & Nearey (2006), Hillenbrand (2012),Morrison (2012)), ou des méthodes de modélisation du signal telles que la transformation cosinoïdale discrète(Harrington 2010) requièrent que soient relevées les valeurs formantiques des voyelles tout au long de leur durée. Sont successivement étudiées la fiabilité de l'extraction automatique, les distributions statistiques des valeurs formantiques de chaque voyelle et les méthodes de normalisation appropriées aux conversations spontanées. Les différences entre les locuteurs sont ensuite évaluées en analysant tour à tour et après normalisation les changements spectraux, les valeurs formantiques à la moitié de la durée de la voyelle et les transformations cosinoïdales. Les méthodes déployées sont les k plus proches voisins, les analyses discriminantes quadratiques et linéaires, ainsi que les régressions linéaires à effets mixtes. Une conclusion temporaire de ce travail est que l'acquisition du contraste/ɪ/-/i:/ semble plus robuste que celle de /ʊ/-/u:/. / This study undertakes to assess the evolution of the phonological acquisition of the English /ɪ/-/i:/ and /ʊ/-/u:/ contrasts by French students. The corpus is made up of recordings of spontaneous conversations with native speakers. 12 students, 9 females and 3 males, were recorded over 4 sessions in six-month intervals. The approach adopted here is resolutely quantitative, and agnostic with respect to theories of second language acquisition such as Flege's, Best's or Kuhl's. In order to assess the potential changes in pronunciations, an automatic procedure of alignment and extraction has been devised, based on PRAAT (Boersma 2001). Phonemic and word alignments had been carried out with SPPAS (Bigi 2012) and P2FA (Yuan & Liberman 2008) beforehand. More than 90,000 vowels were thus collected and analysed. The extracted data consist of information such as the number of syllables in the word, the transcription of its dictionary pronunciation, the structure of the syllable the vowel appears in, of the preceding and succeeding phonemes, their places and manners of articulation, whether they belong to the same word or not, but also especially of the F0, F1, F2, F3 and F4 formant values. These values were collected at each centile of the duration of the vowel, in order to be able to take into account of the influences of consonantal environments. Besides, theories such as vowel-inherent spectral changes (Nearey & Assmann (1986), Morrison & Nearey (2006), Hillenbrand (2012), Morrison (2012)), and methods of signal modelling such as discrete cosine transforms (Harrington 2010) need formant values all throughout the duration of the vowel. Then the reliability of the automatic procedure, the per-vowel statistical distributions of the formant values, and the normalization methods appropriate to spontaneous speech are studied in turn. Speaker differences are assessed by analysing spectral changes, mid-temporal formant values and discrete cosine transforms with normalized values. The methods resorted to are the k nearest neighbours, linear and quadratic discriminant analyses and linear mixed effects regressions. A temporary conclusion is that the acquisition of the /ɪ/-/i:/ contrast seems more robust than that of the /ʊ/-/u:/ contrast.

No description available.