Orientadores: Lee Luan Ling, Fábio Violaro / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-19T05:40:32Z (GMT). No. of bitstreams: 1 GonzalezGonzalez_DianaCristina_M.pdf: 2589944 bytes, checksum: ddbbbef6076eb402f4abe638ebcd232b (MD5) Previous issue date: 2011 / Resumo: Esta dissertação apresenta a implementação de um sistema de Reconhecimento Automático de Locutor (ASR). Este sistema emprega um novo parâmetro de características de locutor baseado no modelo multifractal "VVGM" (Variable Variance Gaussian Multiplier). A metodologia adotada para o desenvolvimento deste sistema foi formulada em duas etapas. Inicialmente foi implementado um sistema ASR tradicional, usando como vetor de características os MFCCs (Mel-Frequency Cepstral Coefficients) e modelo de mistura gaussiana (GMM) como classificador, uma vez que é uma configuração clássica, adotada como referência na literatura. Este procedimento permite ter um conhecimento amplo sobre a produção de sinais de voz, além de um sistema de referência para comparar o desempenho do novo parâmetro VVGM. A segunda etapa foi dedicada ao estudo de processos multifractais em sinais de fala, já que eles enfatizam-se na análise das informações contidas nas partes não estacionárias do sinal avaliado. Aproveitando essa característica, sinais de fala são modelados usando o modelo VVGM. Este modelo é baseado no processo de cascata multiplicativa binomial, e usa as variâncias dos multiplicadores de cada estágio como um novo vetor de característica. As informações obtidas pelos dois métodos são diferentes e complementares. Portanto, é interessante combinar os parâmetros clássicos com os parâmetros multifractais, a fim de melhorar o desempenho dos sistemas de reconhecimento de locutor. Os sistemas propostos foram avaliados por meio de três bases de dados de fala com diferentes configurações, tais como taxas de amostragem, número de falantes e frases e duração do treinamento e teste. Estas diferentes configurações permitem determinar as características do sinal de fala requeridas pelo sistema. Do resultado dos experimentos foi observado que o sistema de identificação de locutor usando os parâmetros VVGM alcançou taxas de acerto significativas, o que mostra que este modelo multifractal contém informações relevantes sobre a identidade de cada locutor. Por exemplo, a segunda base de dados é composta de sinais de fala de 71 locutores (50 homens e 21 mulheres) digitalizados a 22,05 kHz com 16 bits/amostra. O treinamento foi feito com 20 frases para cada locutor, com uma duração total de cerca de 70 s. Avaliando o sistema ASR baseado em VVGM, com locuções de teste de 3 s de comprimento, foi obtida uma taxa de reconhecimento de 91,30%. Usando estas mesmas condições, o sistema ASR baseado em MFCCs atingiu uma taxa de reconhecimento de 98,76%. No entanto, quando os dois parâmetros foram combinados, a taxa de reconhecimento aumentou para 99,43%, mostrando que a nova característica acrescenta informações importantes para o sistema de reconhecimento de locutor / Abstract: This dissertation presents an Automatic Speaker Recognition (ASR) system, which employs a new parameter based on the ¿VVGM? (Variable Variance Gaussian Multiplier) multifractal model. The methodology adopted for the development of this system is formulated in two stages. Initially, a traditional ASR system was implemented, based on the use of Mel-Frequency Cepstral Coefficients (MFCCs) and the Gaussian mixture models (GMMs) as the classifier, since it is the method with the best results in the literature. This procedure allows having a broad knowledge about the production of speech signals and a reference system to compare the performance of the new VVGM parameter. The second stage was dedicated to the study of the multifractal processes for speech signals, given that with them, it is possible to analyze information contained in non-stationary parts of the evaluated signal. Taking advantage of this characteristic, speech signals are modeled using the VVGM model, which is based on the binomial multiplicative cascade process, and uses the variances of multipliers for each state as a new speech feature. The information obtained by the two methods is different and complementary. Therefore, it is interesting to combine the classic parameters with the multifractal parameters in order to improve the performance of speaker recognition systems. The proposed systems were evaluated using three databases with different settings, such as sampling rates, number of speakers and phrases, duration of training and testing. These different configurations allow the determination of characteristics of the speech signal required by the system. With the experiments, the speaker identification system based on the VVGM parameters achieved significant success rates, which shows that this multifractal model contains relevant information of the identity of each speaker. For example, the second database is composed of speech signals of 71 speakers (50 men and 21 women) digitized at 22.05 kHz with 16 bits/sample. The training was done with 20 phrases for each speaker, with an approximately total duration of 70 s. Evaluating the ASR system based on VVGM, with this database and using test locutions with 3s of duration, it was obtained a recognition rate of 91.3%. Using these same conditions, the ASR system based on MFCCs reached a recognition rate of 98.76%. However, when the two parameters are combined, the recognition rate increased to 99.43%, showing that the new feature adds substantial information to the speaker recognition system / Mestrado / Telecomunicações e Telemática / Mestre em Engenharia Elétrica

Thesis (PhD)--Stellenbosch University, 2012. / ENGLISH ABSTRACT: Various properties of integer partitions are studied in this work, in particular the number of summands, the number of ascents and the multiplicities of parts. We work on random partitions, where all partitions from a certain family are equally likely, and determine moments and limiting distributions of the different parameters. The thesis focuses on three main problems: the first of these problems is concerned with the length of prime partitions (i.e., partitions whose parts are all prime numbers), in particular restricted partitions (i.e., partitions where all parts are distinct). We prove a central limit theorem for this parameter and obtain very precise asymptotic formulas for the mean and variance. The second main focus is on the distribution of the number of parts of a given multiplicity, where we obtain a very interesting phase transition from a Gaussian distribution to a Poisson distribution and further to a degenerate distribution, not only in the classical case, but in the more general context of ⋋-partitions: partitions where all the summands have to be elements of a given sequence ⋋ of integers. Finally, we look into another phase transition from restricted to unrestricted partitions (and from Gaussian to Gumbel-distribution) as we study the number of summands in partitions with bounded multiplicities. / AFRIKAANSE OPSOMMING: Verskillende eienskappe van heelgetal-partisies word in hierdie tesis bestudeer, in die besonder die aantal terme, die aantal stygings en die veelvoudighede van terme. Ons werk met stogastiese partisies, waar al die partisies in ’n sekere familie ewekansig is, en ons bepaal momente en limietverdelings van die verskillende parameters. Die teses fokusseer op drie hoofprobleme: die eerste van hierdie probleme gaan oor die lengte van priemgetal-partisies (d.w.s., partisies waar al die terme priemgetalle is), in die besonder beperkte partisies (d.w.s., partisies waar al die terme verskillend is). Ons bewys ’n sentrale limietstelling vir hierdie parameter en verkry baie presiese asimptotiese formules vir die gemiddelde en die variansie. Die tweede hooffokus is op die verdeling van die aantal terme van ’n gegewe veelvoudigheid, waar ons ’n baie interessante fase-oorgang van ’n normaalverdeling na ’n Poisson-verdeling en verder na ’n ontaarde verdeling verkry, nie net in die klassieke geval nie, maar ook in die meer algemene konteks van sogenaamde ⋋-partities: partisies waar al die terme elemente van ’n gegewe ry ⋋ van heelgetalle moet wees.

This thesis is concerned with graphical Gaussian models with equality constraints on the concentration or partial correlation matrix introduced by Højsgaard and Lauritzen (2008) as RCON and RCOR models. The models can be represented by vertex and edge coloured graphs G = (V,ε), where parameters associated with equally coloured vertices or edges are restricted to being identical. In the first part of this thesis we study the problem of estimability of a non-zero model mean μ if the covariance structure Σ is restricted to satisfy the constraints of an RCON or RCOR model but is otherwise unknown. Exploiting results in Kruskal (1968), we obtain a characterisation of suitable linear spaces Ω such that if Σ is restricted as above, the maximum likelihood estimator μ(with circumflex) and the least squares estimator μ* of μ coincide for μ ∈ Ω, thus allowing μ and Σ to be estimated independently. For the special case of Ω being specified by equality relations among the entries of μ according to a partition M of the model variables V, our characterisation translates into a necessary and sufficient regularity condition on M and (V,ε). In the second part we address model selection of RCON and RCOR models. Due to the large number of models, we study the structure of four model classes lying strictly within the sets of RCON and RCOR models, each of which is defined by desirable statistical properties corresponding to colouring regularity conditions. Two of these appear in Højsgaard and Lauritzen (2008), while the other two arise from the regularity condition ensuring equality of estimators μ(with circumflex) = μ* we find in the first part. We show each of the colouring classes to form complete lattices, which qualifies the corresponding model spaces for an Edwards-Havránek model selection procedure (Edwards and Havránek, 1987). We develop a coresponding algorithm for one of the model classes and give an algorithm for a systematic search in accordance with the Edwards-Havránek principles for a second class. Both are applied to data sets previously analysed in the literature, with very encouraging performances.

Salient contour detection is an important lowlevel visual process in the human visual system, and has significance towards understanding higher visual and cognitive processes. Salience detection can be investigated by examining the visual cortical response to visual input. Visual response activity in the early stages of visual processing can be approximated by a sequence of convolutions of the input scene with the difference-of-Gaussian (DoG) and the oriented Gabor filters. The filtered responses are unusually high for prominent edge locations in the image, and are uniformly similar across different natural image inputs. Furthermore, such a response follows a power law distribution. The aim of this thesis is to examine how these response properties could be utilized to the problem of salience detection. First, I identify a method to find the best threshold on the response activity (orientation energy) toward the detection of salient contours: compare the response distribution to a Gaussian distribution of equal variance. Second, I justify this comparison by providing an explanation under the framework of Suspicious Coincidence proposed by Barlow [1]. A connection is provided between perceived salience of contours and the neuronal goal of detecting suspiciousness, where salient contours are seen as affording suspicious coincidences by the visual system. Finally, the neural plausibility of such a salience detection mechanism is investigated, and the representational effciency is shown which could potentially explain why the human visual system can effortlessly detect salience.

Battail [1989] shows that an appropriate criterion for the design of long block codes is the closeness of the normalized weight distribution to a Gaussian distribution. A subsequent work shows that iterated product of single parity check codes satisfy this criterion [1994]. Motivated by these earlier works, in this thesis, we study the effect of the interleaver on the performance of turbo codes for large block lengths, $N\rightarrow\infty$. A parallel concatenated turbo code that consists of two or more component codes is considered. We demonstrate that for $N\rightarrow\infty$, the normalized weight of the systematic $\widehat{w_1}=\displaystyle\frac{w_1}{\sqrt{N}}$, and the parity check sequences $\widehat{w_2}=\displaystyle\frac{w_2}{\sqrt{N}}$ and $\widehat{w_3}=\displaystyle\frac{w_3}{\sqrt{N}}$ become a set of jointly Gaussian distributions for the typical values of $\widehat{w_i},i=1,2,3$, where the typical values of $\widehat{w_i}$ are defined as $\displaystyle\lim_{N\rightarrow\infty}\frac{\widehat{w_i}}{\sqrt{N}}\neq 0,1$ for $i=1,2,3$. To optimize the turbo code performance in the waterfall region which is dominated by high-weight codewords, it is desirable to reduce $\rho_{ij}$, $i,j=1,2,3$ as much as possible, where $\rho_{ij}$ is the correlation coefficient between $\widehat{w_i}$ and $\widehat{w_j}$. It is shown that: (i)~$\rho_{ij}>0$, $i,j=1,2,3$, (ii)~$\rho_{12},\rho_{13}\rightarrow 0$ as $N\rightarrow\infty$, and (iii)~$\rho_{23}\rightarrow 0$ as $N\rightarrow\infty$ for "almost" any random interleaver. This indicates that for $N\rightarrow\infty$, the optimization of the interleaver has a diminishing effect on the distribution of high-weight error events, and consequently, on the error performance in the waterfall region. We show that for the typical weights, this weight distribution approaches the average spectrum defined by Poltyrev [1994]. We also apply the tangential sphere bound (TSB) on the Gaussian distribution in AWGN channel with BPSK signalling and show that it performs very close to the capacity for code rates of interest. We also study the statistical properties of the low-weight codeword structures. We prove that for large block lengths, the number of low-weight codewords of these structures are some Poisson random variables. These random variables can be used to evaluate the asymptotic probability mass function of the minimum distance of the turbo code among all the possible interleavers. We show that the number of indecomposable low-weight codewords of different types tend to a set of independent Poisson random variables. We find the mean and the variance of the union bound in the error floor region and study the effect of expurgating low-weight codewords on the performance. We show that the weight distribution in the transition region between Poisson and Gaussian follows a negative binomial distribution. We also calculate the interleaver gain for multi-component turbo codes based on these Poisson random variables. We show that the asymptotic error performance for multi-component codes in different weight regions converges to zero either exponentially (in the Gaussian region) or polynomially (in the Poisson and negative binomial regions) with respect to the block length, with the code-rate and energy values close to the channel capacity.

Battail [1989] shows that an appropriate criterion for the design of long block codes is the closeness of the normalized weight distribution to a Gaussian distribution. A subsequent work shows that iterated product of single parity check codes satisfy this criterion [1994]. Motivated by these earlier works, in this thesis, we study the effect of the interleaver on the performance of turbo codes for large block lengths, $N\rightarrow\infty$. A parallel concatenated turbo code that consists of two or more component codes is considered. We demonstrate that for $N\rightarrow\infty$, the normalized weight of the systematic $\widehat{w_1}=\displaystyle\frac{w_1}{\sqrt{N}}$, and the parity check sequences $\widehat{w_2}=\displaystyle\frac{w_2}{\sqrt{N}}$ and $\widehat{w_3}=\displaystyle\frac{w_3}{\sqrt{N}}$ become a set of jointly Gaussian distributions for the typical values of $\widehat{w_i},i=1,2,3$, where the typical values of $\widehat{w_i}$ are defined as $\displaystyle\lim_{N\rightarrow\infty}\frac{\widehat{w_i}}{\sqrt{N}}\neq 0,1$ for $i=1,2,3$. To optimize the turbo code performance in the waterfall region which is dominated by high-weight codewords, it is desirable to reduce $\rho_{ij}$, $i,j=1,2,3$ as much as possible, where $\rho_{ij}$ is the correlation coefficient between $\widehat{w_i}$ and $\widehat{w_j}$. It is shown that: (i)~$\rho_{ij}>0$, $i,j=1,2,3$, (ii)~$\rho_{12},\rho_{13}\rightarrow 0$ as $N\rightarrow\infty$, and (iii)~$\rho_{23}\rightarrow 0$ as $N\rightarrow\infty$ for "almost" any random interleaver. This indicates that for $N\rightarrow\infty$, the optimization of the interleaver has a diminishing effect on the distribution of high-weight error events, and consequently, on the error performance in the waterfall region. We show that for the typical weights, this weight distribution approaches the average spectrum defined by Poltyrev [1994]. We also apply the tangential sphere bound (TSB) on the Gaussian distribution in AWGN channel with BPSK signalling and show that it performs very close to the capacity for code rates of interest. We also study the statistical properties of the low-weight codeword structures. We prove that for large block lengths, the number of low-weight codewords of these structures are some Poisson random variables. These random variables can be used to evaluate the asymptotic probability mass function of the minimum distance of the turbo code among all the possible interleavers. We show that the number of indecomposable low-weight codewords of different types tend to a set of independent Poisson random variables. We find the mean and the variance of the union bound in the error floor region and study the effect of expurgating low-weight codewords on the performance. We show that the weight distribution in the transition region between Poisson and Gaussian follows a negative binomial distribution. We also calculate the interleaver gain for multi-component turbo codes based on these Poisson random variables. We show that the asymptotic error performance for multi-component codes in different weight regions converges to zero either exponentially (in the Gaussian region) or polynomially (in the Poisson and negative binomial regions) with respect to the block length, with the code-rate and energy values close to the channel capacity.

Salient contour detection is an important lowlevel visual process in the human visual system, and has significance towards understanding higher visual and cognitive processes. Salience detection can be investigated by examining the visual cortical response to visual input. Visual response activity in the early stages of visual processing can be approximated by a sequence of convolutions of the input scene with the difference-of-Gaussian (DoG) and the oriented Gabor filters. The filtered responses are unusually high for prominent edge locations in the image, and are uniformly similar across different natural image inputs. Furthermore, such a response follows a power law distribution. The aim of this thesis is to examine how these response properties could be utilized to the problem of salience detection. First, I identify a method to find the best threshold on the response activity (orientation energy) toward the detection of salient contours: compare the response distribution to a Gaussian distribution of equal variance. Second, I justify this comparison by providing an explanation under the framework of Suspicious Coincidence proposed by Barlow [1]. A connection is provided between perceived salience of contours and the neuronal goal of detecting suspiciousness, where salient contours are seen as affording suspicious coincidences by the visual system. Finally, the neural plausibility of such a salience detection mechanism is investigated, and the representational effciency is shown which could potentially explain why the human visual system can effortlessly detect salience.

This thesis investigates the design and implementation strategies for a Taiwanese speech recognition system. It adopts a 4 plus 1¡]five times¡^recording strategy, where the 1st four recordings are used for speech feature training and the last recording for speech recognition simulation. Mel-frequency cepstrum coefficients and hidden Markov model are used as the feature model and the recognition model respectively. Under the Intel Celeron 2.4 GHz personal computer and Red Hat Linux 9.0 operating system environment, a correct phrase recognition rate of 90% can be reached for a 4200 Taiwanese phrase database.

Tez (Yüksek Lisans) - Süleyman Demirel Üniversitesi, Fen Bilimleri Enstitüsü, Fizik Anabilim Dalı, 2008. / Kaynakça var.

Thesis (Ph. D.)--University of California, Riverside, 2009. / Includes abstract. Also issued in print. Includes bibliographical references (leaves 81-84). Available via ProQuest Digital Dissertations.