In the present day, the music becomes more popular due to the following three reasons: (1) the evolution of the MP3 compression technology, (2) the growth of the public platform, and (3) the development of the MP3 portable discs. Most people follow the music to hum or follow the rhythm to tap sometimes. The meanings of a music style may be various if it is explained or felt by different people. Therefore we cannot obtain a very explicit answer if the notation of the music cannot be exactly made. We need some techniques and methods to analyze the music, and obtain some of its embedded information. Tempo and beats are very important elements in the perceptual music. Therefore, tempo estimation and beat tracking are fundamental techniques in automatic audio processing, which are crucial to multimedia applications. In this thesis, we first develop an artificial neural network to classify the music excerpts into the evaluation preference. And then, with the preference classification, we can obtain accurate estimation for tempo and beats, by either Ellis's method or Dixon's method. We test our method with a mixed data set which contains ten music genres extracted from the "ballroom dancer" database. Our experimental results show that the accuracy of our method is higher than that of only one individual Ellis's method or Dixon's method.

An attempt was made in this study to classify copper deposits using the ratio of copper, gold and silver in the deposits. Data were accumulated from a literature search and neutron activation analysis of samples. Triangular diagrams were plotted and fields for each type of deposit were defined. Porphyry, volcanogenic, and sedimentary fields were most easily defined due to the amount and quality of data available for these deposits. It is found that the size of the field for any type of deposit is a function of its mineralogy. This is illustrated by the data spread for the copper and lead-zinc zones in volcanogenic deposits. A clear separation of deposit types could not be obtained due to a partial overlap of the data fields; however, it is suggested that further work using more metals might yield a clear separation of deposits. / Thesis / Bachelor of Science (BSc)

The main purpose of the study is to compare the differences of marketing strategies of the two-dimension classification, which classified by degree of customization and benefit duration period. And to explore the influences of the strategic selection of the services industries by different organization¡¦s features and different strategies maker¡¦s characters. The research takes eight industries as population to investigate their marketing strategies. After collected, the data is processed with frequency analysis, factor analysis, discriminant analysis, one-way ANOVA, and two-way ANOVA. The key findings are: 1. The result indicates that two dimension classification method is more efficiently to reflect the differences of the marketing strategic between different taxonomies of service. 2. Service of ¡§high degree of customization and long benefit duration period¡¨ focuses on the marketing strategies such as ¡§public relationship strategy,¡¨ ¡§strategy of emphasizing the importance of employees,¡¨ and ¡§building the intangible quality image strategy, The strategies such as ¡§low price strategy,¡¨ ¡§advertisement strategy,¡¨ and ¡§non-personal promotion,¡¨ which only can bringing short-term effect, are hardly used. 3. Service of ¡§low degree of customization and short benefit duration period¡¨ is just the opposite, emphasizing the strategies which can bring the instant effect. 4. Service of ¡§high degree of customization and short benefit duration period¡¨ prefers the use of ¡§strategy under smoothing the unbalance of supply and demand.¡¨

Recent technique advances have made sensor nodes to be smaller, cheaper and more powerful. Compared with traditional centralized sensing systems, wireless sensor networks are very easy to deploy and can be deployed densely. They have a better sensing coverage and provide more reliable information delivery. Those advantages make wireless sensor networks very useful in a wide variety of applications. As one of active research areas, acoustic monitoring with wireless sensor networks is still new, and very few applications can recognize human voice, discriminate human speech and music, or identify individual speakers. In this thesis work, we designed and implemented an acoustic monitoring system with a wireless sensor network to classify human voice versus music. We also introduce a new, effective sound source localization method, using Root Mean Square (RMS) detected by different nodes of a wireless sensor network to estimate the speaker's location. The experimental results show that our approaches are effective. This research could form a basis for further developing speech recognition, speaker identification, even emotion detection with wireless sensor networks.

在這科技飛騰的時代，智慧型手機使用日益普及，根據eMarketer於2016年公布台灣使用智慧型手機人口佔總人口73.4%，相較於新加坡71.8%與南韓70.4%的使用率，此比率高居全球之冠，各行業該如何運用智慧型手機市場為自己駐足的行業開創佳績，成為現今人們廣為關注的話題。 本論文研究所用之資料取自「科技部傳播調查資料庫第一期第三次（2014）：媒體的娛樂與社交功能」一般民眾（18 歲以上）之問卷資料。首先對樣本基本資料結構與特性進行描述，接著將智慧型手機使用的相關題項找出，並進行因素分析找出因素構面作為分群變數，藉由兩階段分群法進行分群，探討其各群間相關之特性與智慧型手機使用之目的。爾後從性別、年齡與教育程度等基本人口變項進行分析，進一步了解不同人口基本結構智慧型手機之使用目的之差異情形，並將「網路素養」、「社交媒體」等相關題組進行因素分析，萃取出重要共同因素後並予以命名，以探討不同媒體社交功能使用情形與智慧型手機使用目的之相關性，最後將人口基本結構與共同因素視為變數，分別採用CART、C5.0、QUEST與CHAID四種決策樹分析方法對「集群一」、「集群二」智慧型手機高度使用者進行模型之建構，使各行業可針對欲探討之集群提出行銷方針。 / With the rapid development of technology, the Internet and mobile phones play an important role in our lives. According to eMarketer 2016, 73.4% of Taiwan's population use smartphones, compared to 71.8% in Singapore and 70.4% in South Korea , Taiwan tops the list of the world. How to create success by using smartphone market is an important issue today. The data used in this thesis was taken from the Ministry of Science and Technology Survey in 2014. The survey topic was media entertainment and social functions, based on general public who are 18 years old or older. First, the structures of the sample are described. Next, we extract factors by using factor analysis. The factors are used as the cluster variables. This study uses two-stage method to cluster and explore characteristics of the relevant groups for the smartphone usage. Then, we analyze demographic variables to understand different populations of smart phones usage, and extract common factors of "Internet Literacy" and "Social Media" by using factor analysis. Finally, the basic structure of the population and the common factors are used to classify smartphone users, which helps to provide marketing guidelines.

The theme of politics has been popular among authors of African language literary works in the past years. In Setswana literature authors often explore this theme in dramas. Among those scholars who have discussed political themes in Setswana dramas are K.Mogapi (1985) and J.S.Shole (1988). The aim of this dissertation is to analyze politics in Setswana dramatic works. This mini-dissertation uses three concepts, namely defininition, interpretation and classification in its analysis of Setswana dramas. Furthermore, this dissertation has selected an adapted narratological model, which stresses topic as highly significant in understanding the content level of a text, as it links the events of the content coherently. This model also highlights theme as the most important aspect of the compositional level, as it links the events of the plot. The adapted narratological models also emphasises atmosphere when discussing style. These three levels are important when discussing politics in Setswana drama. Because this research investigates political drama, a few concepts such as politics and society, politics and democracy, and politics and literature are explained. The main aim of this dissertation is to analyze politics in Setswana dramatic works. This mini-dissertation discusses the theme of politics in Setswana drama according to three selected periods, namely (a) the period 1930-1993, which is represented by the drama MotswaseleII (1945) by L.D.Raditladi, (b) the period 1994-1995, which is represented by the drama Kaine le Abele (1995) by G.Mokae and (c) the period 1996-2002, which is represented by the drama Diterama tsa ga Zakes Mda (2002) translated into Setswana by P.M.Sebate. The results of the investigation can be summarized as follows: During the period 1930-1993, authors tend to write about traditional politics, that is, chieftainship, which is passed on by birthright. In his drama MotswaseleII, Raditladi uses a number of techniques in the development of his message of looking down upon traditional governance or leadership. During the period 1994-1995, which is represented by the drama Kaine le Abele, Setswana drama deals with modern or contemporary politics, and tends to show or depict the cruelty of the then apartheid South African government. While examining the period 1996-2002, which is represented by a collection of dramas by Mda, three short dramas were selected. In the drama ‘Re tla opelela lefatshe la borrarona’, prominent techniques are rhetoric question, motif, flashback and contrast. At the plot level of the drama ‘Mantswe a lefifi a a lela’, Mda deployed several prominent techniques to further the development of the theme of politics in his text, but this dissertation has selected only two main techniques, which are tragedy and ellision. In the drama ‘Tsela’, the author has used a number of techniques, but this dissertation has selected two pronounced techniques, which are complication of events and symbolism. Lastly, the findings of this research demonstrate the usefulness of the classification of Setswana political dramas, written up to now, according to three periods, namely (a) the period 1930-1993, (b) the period 1994-1995, and (c) the period 1996-2002. / Dissertation (MA (African Languages))--University of Pretoria, 2007. / African Languages / unrestricted

Mechanisms for charge transport in organic electronics allows them to perform with disordered internal morphology, something which is not possible for traditional crystalline semiconductors. Improvements to performance can occur when the materials change their relative positions to each other, resulting as a different spatial dispersion with lower electrical loss over the device area. A numerical method has been developed using interaction models for molecules from colloidal self-assembly. Colloids are rigid particles with a volume which is embodied by their shape and their collective behaviour depends on its density. The self-assembly mechanism used is condensation, which increases the density by removing the spaces between molecules while they lose thermal energy due to the increasing steric interactions with neighbours. The molecular chemical structure determines the spatial probability of electron orbitals that (for a given energy) outlines their geometric shape. Because these shapes are localized onto the molecule, their intermolecular positions determine how close these orbitals can be to each other which is important for electron charge transport. During operation, the organic active layer may have thermal energy to cause molecular reorganization before cooling, which increases the probability to find disordered states within the device. A comprehensive suite of tools has been developed which can classify disorder in the physical characteristics of morphology; such as density, internal spacing, and angular orientation symmetry. These tools where used to optimize the experimental preparations for depositing nanoparticle dispersions on surfaces within organic electronic devices. These have also been used to quantify the statistical variations in structure between configurations produced from our Monte Carlo method and a similar molecular dynamics approach. Simulated self-assembly within highly confined areas showed repeatedly sampled microstates, suggesting that at thermodynamic equilibrium confined particles have quantized density states. We conclude with morphologies resulting from non-circular shapes and systems of donor-acceptor type molecules. / Thesis / Doctor of Philosophy (PhD)

É de conhecimento geral que, em alguns casos, as classes são espectralmente muito similares e que não é possível separá-las usando dados convencionais em baixa dimensionalidade. Entretanto, estas classes podem ser separáveis com um alto grau de acurácia em espaço de alta dimensão. Por outro lado, classificação de dados em alta dimensionalidade pode se tornar um problema para classificadores paramétricos, como o Máxima Verossimilhança Gaussiana (MVG). Um grande número de variáveis que caracteriza as imagens hiperespectrais resulta em um grande número de parâmetros a serem estimados e, geralmente, tem-se um número limitado de amostras de treinamento disponíveis. Essa condição causa o fenômeno de Hughes que consiste na gradual degradação da acurácia com o aumento da dimensionalidade dos dados. Neste contexto, desperta o interesse a utilização de classificadores não-paramétricos, como é o caso de Support Vector Machines (SVM). Nesta dissertação é analisado o desempenho do classificador SVM quando aplicado a imagens hiperespectrais de sensoriamento remoto. Inicialmente os conceitos teóricos referentes à SVM são revisados e discutidos. Em seguida, uma série de experimentos usando dados AVIRIS são realizados usando diferentes configurações para o classificador. Os dados cobrem uma área de teste da Purdue University e apresenta classes de culturas agrícolas espectralmente muito similares. A acurácia produzida na classificação por diferentes kernels são investigadas em função da dimensionalidade dos dados e comparadas com as obtidas com o classificador MVG. Como SVM é aplicado a um par de classes por vez, desenvolveu-se um classificador multi-estágio estruturado em forma de árvore binária para lidar como problema multi-classe. Em cada nó, a seleção do par de classes mais separáveis é feita pelo critério distância de Bhattacharyya. Tais classes darão origem aos nós descendentes e serão responsáveis por definir a função de decisão SVM. Repete-se este procedimento em todos os nós da árvore, até que reste apenas uma classe por nó, nos chamados nós terminais. Os softwares necessários foram desenvolvidos em ambiente MATLAB e são apresentados na dissertação. Os resultados obtidos nos experimentos permitem concluir que SVM é uma abordagem alternativa válida e eficaz para classificação de imagens hiperespectrais de sensoriamento remoto. / This dissertation deals with the application of Support Vector Machines (SVM) to the classification of remote sensing high-dimensional image data. It is well known that in many cases classes that are spectrally very similar and thus not separable when using the more conventional low-dimensional data, can nevertheless be separated with an high degree of accuracy in high dimensional spaces. Classification of high-dimensional image data can, however, become a challenging problem for parametric classifiers such as the well-known Gaussian Maximum Likelihood. A large number of variables produce an also large number of parameters to be estimated from a generally limited number of training samples. This condition causes the Hughes phenomenon which consists in a gradual degradation of the accuracy as the data dimensionality increases beyond a certain value. Non-parametric classifiers present the advantage of being less sensitive to this dimensionality problem. SVM has been receiving a great deal of attention from the international community as an efficient classifier. In this dissertation it is analyzed the performance of SVM when applied to remote sensing hyper-spectral image data. Initially the more theoretical concepts related to SVM are reviewed and discussed. Next, a series of experiments using AVIRIS image data are performed, using different configurations for the classifier. The data covers a test area established by Purdue University and presents a number of classes (agricultural fields) which are spectrally very similar to each other. The classification accuracy produced by different kernels is investigated as a function of the data dimensionality and compared with the one yielded by the well-known Gaussian Maximum Likelihood classifier. As SVM apply to a pair of classes at a time, a multi-stage classifier structured as a binary tree was developed to deal with the multi-class problem. The tree classifier is initially defined by selecting at each node the most separable pair of classes by using the Bhattacharyya distance as a criterion. These two classes will then be used to define the two descending nodes and the corresponding SVM decision function. This operation is performed at every node across the tree, until the terminal nodes are reached. The required software was developed in MATLAB environment and is also presented in this dissertation.

É de conhecimento geral que, em alguns casos, as classes são espectralmente muito similares e que não é possível separá-las usando dados convencionais em baixa dimensionalidade. Entretanto, estas classes podem ser separáveis com um alto grau de acurácia em espaço de alta dimensão. Por outro lado, classificação de dados em alta dimensionalidade pode se tornar um problema para classificadores paramétricos, como o Máxima Verossimilhança Gaussiana (MVG). Um grande número de variáveis que caracteriza as imagens hiperespectrais resulta em um grande número de parâmetros a serem estimados e, geralmente, tem-se um número limitado de amostras de treinamento disponíveis. Essa condição causa o fenômeno de Hughes que consiste na gradual degradação da acurácia com o aumento da dimensionalidade dos dados. Neste contexto, desperta o interesse a utilização de classificadores não-paramétricos, como é o caso de Support Vector Machines (SVM). Nesta dissertação é analisado o desempenho do classificador SVM quando aplicado a imagens hiperespectrais de sensoriamento remoto. Inicialmente os conceitos teóricos referentes à SVM são revisados e discutidos. Em seguida, uma série de experimentos usando dados AVIRIS são realizados usando diferentes configurações para o classificador. Os dados cobrem uma área de teste da Purdue University e apresenta classes de culturas agrícolas espectralmente muito similares. A acurácia produzida na classificação por diferentes kernels são investigadas em função da dimensionalidade dos dados e comparadas com as obtidas com o classificador MVG. Como SVM é aplicado a um par de classes por vez, desenvolveu-se um classificador multi-estágio estruturado em forma de árvore binária para lidar como problema multi-classe. Em cada nó, a seleção do par de classes mais separáveis é feita pelo critério distância de Bhattacharyya. Tais classes darão origem aos nós descendentes e serão responsáveis por definir a função de decisão SVM. Repete-se este procedimento em todos os nós da árvore, até que reste apenas uma classe por nó, nos chamados nós terminais. Os softwares necessários foram desenvolvidos em ambiente MATLAB e são apresentados na dissertação. Os resultados obtidos nos experimentos permitem concluir que SVM é uma abordagem alternativa válida e eficaz para classificação de imagens hiperespectrais de sensoriamento remoto. / This dissertation deals with the application of Support Vector Machines (SVM) to the classification of remote sensing high-dimensional image data. It is well known that in many cases classes that are spectrally very similar and thus not separable when using the more conventional low-dimensional data, can nevertheless be separated with an high degree of accuracy in high dimensional spaces. Classification of high-dimensional image data can, however, become a challenging problem for parametric classifiers such as the well-known Gaussian Maximum Likelihood. A large number of variables produce an also large number of parameters to be estimated from a generally limited number of training samples. This condition causes the Hughes phenomenon which consists in a gradual degradation of the accuracy as the data dimensionality increases beyond a certain value. Non-parametric classifiers present the advantage of being less sensitive to this dimensionality problem. SVM has been receiving a great deal of attention from the international community as an efficient classifier. In this dissertation it is analyzed the performance of SVM when applied to remote sensing hyper-spectral image data. Initially the more theoretical concepts related to SVM are reviewed and discussed. Next, a series of experiments using AVIRIS image data are performed, using different configurations for the classifier. The data covers a test area established by Purdue University and presents a number of classes (agricultural fields) which are spectrally very similar to each other. The classification accuracy produced by different kernels is investigated as a function of the data dimensionality and compared with the one yielded by the well-known Gaussian Maximum Likelihood classifier. As SVM apply to a pair of classes at a time, a multi-stage classifier structured as a binary tree was developed to deal with the multi-class problem. The tree classifier is initially defined by selecting at each node the most separable pair of classes by using the Bhattacharyya distance as a criterion. These two classes will then be used to define the two descending nodes and the corresponding SVM decision function. This operation is performed at every node across the tree, until the terminal nodes are reached. The required software was developed in MATLAB environment and is also presented in this dissertation.

É de conhecimento geral que, em alguns casos, as classes são espectralmente muito similares e que não é possível separá-las usando dados convencionais em baixa dimensionalidade. Entretanto, estas classes podem ser separáveis com um alto grau de acurácia em espaço de alta dimensão. Por outro lado, classificação de dados em alta dimensionalidade pode se tornar um problema para classificadores paramétricos, como o Máxima Verossimilhança Gaussiana (MVG). Um grande número de variáveis que caracteriza as imagens hiperespectrais resulta em um grande número de parâmetros a serem estimados e, geralmente, tem-se um número limitado de amostras de treinamento disponíveis. Essa condição causa o fenômeno de Hughes que consiste na gradual degradação da acurácia com o aumento da dimensionalidade dos dados. Neste contexto, desperta o interesse a utilização de classificadores não-paramétricos, como é o caso de Support Vector Machines (SVM). Nesta dissertação é analisado o desempenho do classificador SVM quando aplicado a imagens hiperespectrais de sensoriamento remoto. Inicialmente os conceitos teóricos referentes à SVM são revisados e discutidos. Em seguida, uma série de experimentos usando dados AVIRIS são realizados usando diferentes configurações para o classificador. Os dados cobrem uma área de teste da Purdue University e apresenta classes de culturas agrícolas espectralmente muito similares. A acurácia produzida na classificação por diferentes kernels são investigadas em função da dimensionalidade dos dados e comparadas com as obtidas com o classificador MVG. Como SVM é aplicado a um par de classes por vez, desenvolveu-se um classificador multi-estágio estruturado em forma de árvore binária para lidar como problema multi-classe. Em cada nó, a seleção do par de classes mais separáveis é feita pelo critério distância de Bhattacharyya. Tais classes darão origem aos nós descendentes e serão responsáveis por definir a função de decisão SVM. Repete-se este procedimento em todos os nós da árvore, até que reste apenas uma classe por nó, nos chamados nós terminais. Os softwares necessários foram desenvolvidos em ambiente MATLAB e são apresentados na dissertação. Os resultados obtidos nos experimentos permitem concluir que SVM é uma abordagem alternativa válida e eficaz para classificação de imagens hiperespectrais de sensoriamento remoto. / This dissertation deals with the application of Support Vector Machines (SVM) to the classification of remote sensing high-dimensional image data. It is well known that in many cases classes that are spectrally very similar and thus not separable when using the more conventional low-dimensional data, can nevertheless be separated with an high degree of accuracy in high dimensional spaces. Classification of high-dimensional image data can, however, become a challenging problem for parametric classifiers such as the well-known Gaussian Maximum Likelihood. A large number of variables produce an also large number of parameters to be estimated from a generally limited number of training samples. This condition causes the Hughes phenomenon which consists in a gradual degradation of the accuracy as the data dimensionality increases beyond a certain value. Non-parametric classifiers present the advantage of being less sensitive to this dimensionality problem. SVM has been receiving a great deal of attention from the international community as an efficient classifier. In this dissertation it is analyzed the performance of SVM when applied to remote sensing hyper-spectral image data. Initially the more theoretical concepts related to SVM are reviewed and discussed. Next, a series of experiments using AVIRIS image data are performed, using different configurations for the classifier. The data covers a test area established by Purdue University and presents a number of classes (agricultural fields) which are spectrally very similar to each other. The classification accuracy produced by different kernels is investigated as a function of the data dimensionality and compared with the one yielded by the well-known Gaussian Maximum Likelihood classifier. As SVM apply to a pair of classes at a time, a multi-stage classifier structured as a binary tree was developed to deal with the multi-class problem. The tree classifier is initially defined by selecting at each node the most separable pair of classes by using the Bhattacharyya distance as a criterion. These two classes will then be used to define the two descending nodes and the corresponding SVM decision function. This operation is performed at every node across the tree, until the terminal nodes are reached. The required software was developed in MATLAB environment and is also presented in this dissertation.