Orientador: Reginaldo José Donatelli / Resumo: A variação vocal é um dos processos que pode levar a especiação em aves. A divergência das características vocais pode levar ao isolamento reprodutivo, uma vez que indivíduos de uma mesma espécie não seriam capazes de reconhecer seus pares que emitam sinais vocais alterados. Sabendo-se que o ambiente exerce uma pressão seletiva importante na transmissão do som, os indivíduos tendem a alterar a estrutura acústica de suas vocalizações, aprimorando a transmissão sonora. Assim, indivíduos de uma mesma espécie que habitam ambientes distintos tendem a apresentar diferenças vocais. Este estudo analisou as vocalizações de T. major oriundas de diversas localidades da região Neotropical para identificar possíveis variações vocais entre as diversas subespécies. Estas variações foram correlacionadas com as coordenadas geográficas, com altitude e com a distribuição espacial das subespécies de T. major. Foram coletadas 119 vocalizações de indivíduos distintos, que tiveram as suas variáveis temporais e espectrais analisadas por meio do espectrograma e do espectro de amplitude, tomando um limite de -42 dB como referência. Tais características foram correlacionadas com as variáveis de latitude, longitude e altitude. Ao total, foram obtidas quatro subespécies. As variáveis foram reduzidas por uma análise de Componente Principal, e depois classificadas por uma Análise de Função Discriminante. A partir das correlações, observou-se uma relação inversa entre características espectrais e temporais.... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: The vocal variation is one of the processes that can lead to speciation in birds. The divergence of the vocal features can drive the reproductive isolation, since individuals of the same species will not be able to recognize their mates, which emit altered voice signals. The environment make a significant selective pressure on sound transmission. Thus, individuals tend to change the acoustic structure of their vocalizations to improve the sound transmission. Individuals of the same species that inhabit different environments tend to display vocal differences. This study surveyed the vocalizations of T. major from several locations in the Neotropical region to identify possible vocal variations between different subspecies. These variations were correlated with the geographic coordinates, with altitude and with the spatial distribution of subspecies. We collected 119 vocalizations of different individuals. Vocalizations had their temporal and spectral variables measured through the espectrogram and the power spectrum, taking a -42 dB as reference limit. These variables were correlated with latitude, longitude and altitude. From these same data, we identified the predominant subspecies the location of sampling. In total, four subspecies were obtained. The variables were reduced by a principal component analysis. Then, they were classified by a discriminant function analysis. From the correlations, we found an inverse relationship between spectral and temporal characteristics. F... (Complete abstract click electronic access below) / Mestre

Submitted by GUILHERME SEMENTILI CARDOSO null (guisemcar@yahoo.com.br) on 2016-04-26T18:35:28Z No. of bitstreams: 1 Dissertação_Guilherme_biblioteca.docx: 8022519 bytes, checksum: f94b5576b996645e7a83c64cb61c053c (MD5) / Rejected by Felipe Augusto Arakaki (arakaki@reitoria.unesp.br), reason: Solicitamos que realize uma nova submissão seguindo as orientações abaixo: A versão final da dissertação/tese deve ser submetida no formato PDF (Portable Document Format). O arquivo PDF não deve estar protegido e a dissertação/tese deve estar em um único arquivo, inclusive os apêndices e anexos, se houver. Por favor, corrija o formato do arquivo e realize uma nova submissão. Agradecemos a compreensão. on 2016-04-28T18:58:02Z (GMT) / Submitted by GUILHERME SEMENTILI CARDOSO null (guisemcar@yahoo.com.br) on 2016-05-02T17:21:10Z No. of bitstreams: 1 Dissertação_Guilherme_biblioteca.pdf: 3396397 bytes, checksum: d445e38209f4ffe6e3bc7cb908f14709 (MD5) / Approved for entry into archive by Felipe Augusto Arakaki (arakaki@reitoria.unesp.br) on 2016-05-04T18:22:45Z (GMT) No. of bitstreams: 1 cardoso_gs_me_bot.pdf: 3396397 bytes, checksum: d445e38209f4ffe6e3bc7cb908f14709 (MD5) / Made available in DSpace on 2016-05-04T18:22:45Z (GMT). No. of bitstreams: 1 cardoso_gs_me_bot.pdf: 3396397 bytes, checksum: d445e38209f4ffe6e3bc7cb908f14709 (MD5) Previous issue date: 2016-02-09 / Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) / A variação vocal é um dos processos que pode levar a especiação em aves. A divergência das características vocais pode levar ao isolamento reprodutivo, uma vez que indivíduos de uma mesma espécie não seriam capazes de reconhecer seus pares que emitam sinais vocais alterados. Sabendo-se que o ambiente exerce uma pressão seletiva importante na transmissão do som, os indivíduos tendem a alterar a estrutura acústica de suas vocalizações, aprimorando a transmissão sonora. Assim, indivíduos de uma mesma espécie que habitam ambientes distintos tendem a apresentar diferenças vocais. Este estudo analisou as vocalizações de T. major oriundas de diversas localidades da região Neotropical para identificar possíveis variações vocais entre as diversas subespécies. Estas variações foram correlacionadas com as coordenadas geográficas, com altitude e com a distribuição espacial das subespécies de T. major. Foram coletadas 119 vocalizações de indivíduos distintos, que tiveram as suas variáveis temporais e espectrais analisadas por meio do espectrograma e do espectro de amplitude, tomando um limite de -42 dB como referência. Tais características foram correlacionadas com as variáveis de latitude, longitude e altitude. Ao total, foram obtidas quatro subespécies. As variáveis foram reduzidas por uma análise de Componente Principal, e depois classificadas por uma Análise de Função Discriminante. A partir das correlações, observou-se uma relação inversa entre características espectrais e temporais. Valores de frequência decrescem com o aumento da latitude, enquanto os valores temporais aumentam. Do mesmo modo, os valores de frequência tendem a aumentar com o acréscimo da longitude, enquanto os valores temporais decrescem. Algo semelhante ocorre com os grupos atribuídos às subespécies, pois aquelas amostras que estão mais a Noroeste (T. m. semifasciatus e T. m. melanurus) apresentam menores frequências e maiores durações, enquanto aquelas que se situam mais a sudeste (T. m. major e T. m. stagurus) apresentam maiores frequências e menores durações. Este efeito pode estar vinculado às características do hábitat, pois as subespécies T. m. semifasciatus e T. m. melanurus estão situadas em regiões de predomínio da Floresta Ombrófila Densa da bacia Amazônica. Deste modo, vocalizações com menores valores de frequência e maiores valores de tempo sofrem menor atenuação sonora que seria causada pela alta densidade de obstáculos para a transmissão. Este tipo de variação vocal é bem documentado, sendo suportado pela Hipótese da Adaptação Acústica. / The vocal variation is one of the processes that can lead to speciation in birds. The divergence of the vocal features can drive the reproductive isolation, since individuals of the same species will not be able to recognize their mates, which emit altered voice signals. The environment make a significant selective pressure on sound transmission. Thus, individuals tend to change the acoustic structure of their vocalizations to improve the sound transmission. Individuals of the same species that inhabit different environments tend to display vocal differences. This study surveyed the vocalizations of T. major from several locations in the Neotropical region to identify possible vocal variations between different subspecies. These variations were correlated with the geographic coordinates, with altitude and with the spatial distribution of subspecies. We collected 119 vocalizations of different individuals. Vocalizations had their temporal and spectral variables measured through the espectrogram and the power spectrum, taking a -42 dB as reference limit. These variables were correlated with latitude, longitude and altitude. From these same data, we identified the predominant subspecies the location of sampling. In total, four subspecies were obtained. The variables were reduced by a principal component analysis. Then, they were classified by a discriminant function analysis. From the correlations, we found an inverse relationship between spectral and temporal characteristics. Frequency values decrease with increasing latitude, while time values increase. Moreover, the frequency values tend to increase with the increase in longitude, while time values decrease. Something similar occurs with the groups assigned to the subspecies, for the northwestern samples (attributed to T. m. semifasciatus and T. m. melanurus) have lower frequencies and longer durations, while the southeastern ones (attributed to T. m. major and T. m. stagurus) have higher frequencies and shorter durations. This effect is linked to habitat features, since the subspecies T. m. semifasciatus and T. m. melanurus are located predominantly in regions of dense rain forest of the Amazon basin. Thus, vocalizations with lower frequency values and greater time values suffer less sound attenuation caused by the high density of obstacles to transmission. This kind of vocal variation is well documented, and supported by the Acoustic Adaptation Hypothesis. / CNPq: 133992/2014-4

Speech recognition systems have been developed for the major languages of the world, yet for the majority of languages there are currently no large vocabulary continuous speech recognition (LVCSR) systems. The development of an LVCSR system for a new language is very costly, mainly because a large speech database has to be compiled to robustly capture the acoustic characteristics of the new language. This thesis investigates techniques that enable the re-use of acoustic information from a source language, in which a large amount of data is available, in implementing a system for a new target language. The assumption is that too little data is available in the target language to train a robust speech recognition system on that data alone, and that use of acoustic information from a source language can improve the performance of a target language recognition system. Strategies for cross-language use of acoustic information are proposed, including training on pooled source and target language data, adaptation of source language models using target language data, adapting multilingual models using target language data and transforming source language data to augment target language data for model training. These strategies are allied with Bayesian and transformation-based techniques, usually used for speaker adaptation, as well as with discriminative learning techniques, to present a framework for cross-language re-use of acoustic information. Extensions to current adaptation techniques are proposed to improve the performance of these techniques specifically for cross-language adaptation. A new technique for transformation-based adaptation of variance parameters and a cost-based extension of the minimum classification error (MCE) approach are proposed. Experiments are performed for a large number of approaches from the proposed framework for cross-language re-use of acoustic information. Relatively large amounts of English speech data are used in conjunction with smaller amounts of Afrikaans speech data to improve the performance of an Afrikaans speech recogniser. Results indicate that a significant reduction in word error rate (between 26% and 50%, depending on the amount of Afrikaans data available) is possible when English acoustic data is used in addition to Afrikaans speech data from the same database (i.e both sets of data were recorded under the same c`12onditions and the same labelling process was used). For same-database experiments, best results are achieved for approaches that train models on pooled source and target language data and then perform further adaptation of the models using Bayesian or discriminative techniques on target language data only. Experiments are also performed to evaluate the use of English data from a different database than the Afrikaans data. Peak reductions in word error rate of between 16% and 35% are delivered, depending on the amount of Afrikaans data available. Best results are achieved for an approach that performs a simple transformation of source model parameters using target language data, and then performs Bayesian adaptation of the transformed model on target language data. / Thesis (PhD (Electrical, Electronic and Computer Engineering))--University of Pretoria, 2006. / Electrical, Electronic and Computer Engineering / unrestricted

Several animals, both vertebrate and invertebrate, communicate using sound and they do so in a non-ideal medium, the habitat in which they live. As acoustic signals pass through the habitat, they suffer loss of information due to signal degradation, which is often determined by the acoustic properties of the habitat. Understanding the influence of habitat-induced constraints on signaling is vital to the understanding of evolution of signal structure. Over time signals can evolve their temporal and/or spectral characteristics so as to resist or reduce degradation. Conversely, signalers may modify their behaviour so as to improve transmission for effective communication. The Acoustic Adaptation Hypothesis (AAH) predicts that given the properties of the habitat in which animals communicate, signals should evolve so as to maximize their broadcast range and the number of potential receivers. The prediction of the hypothesis is that signal transmission is best in the native habitat as compared to non-native habitats. Since its inception, the idea of acoustic adaptation has been tested in several vertebrate species including birds, mammals and frogs but rarely in invertebrates. Moreover, most of these studies have been carried out at the macrohabitat level, such as transmission in forests versus grasslands. For animals with limited mobility such as invertebrates, a finer level of investigation at the microhabitat level is more relevant. Only one study on cicadas has investigated the predictions of the AAH at the microhabitat level. Besides, investigations done at the community level are largely missing in the literature. The effect of height on signal transmission is well documented, however, only one study on birds has investigated the AAH with respect to vertical stratification of coexisting species. Among invertebrates, crickets are well known for their calling songs which males use to attract conspecific females over long distances. No study so far has tested the prediction of the AAH in these acoustically communicating invertebrates. The central focus of this study was to test the predictions of the AAH in a natural assemblage of ensiferan (cricket) species. I examined the prediction of the hypothesis at the microhabitat level with respect to the vertical stratification of co-existing ensiferan species. The study was carried out on an assemblage of crickets in the wet evergreen forests of Kudremukh National Park in the Western Ghats in Southern India. For this purpose, it was important to examine calling height and microhabitat selection in these animals because if the use of calling height and microhabitat was random, then there would not be any native height/microhabitat and the question of acoustic adaption would not arise. Therefore, I first standardized methods to characterize the habitat of the crickets. Using resource selection functions, I then quantified microhabitat selection in 13 ensiferan species. I also examined the calling heights of these species. My results suggest that these species are microhabitat specialists and also distribute vertically within the forest with respect to calling height. Based on the knowledge of the vertical distribution of these animals in the forest I then carried out playback experiments using natural calls of 12 species of Ensifera in their natural habitat. The transmission experiments were carried out at five heights in the forest, including the ground, different parts of the understorey as well as in the canopy. The study aimed to examine whether vertical stratification in the ensiferan species of Kudremukh is driven by selection for maximizing transmission range. I examined the effect of different heights on signal degradation. The investigation was carried out with respect to three different measures of signal degradation, namely, total attenuation, signal to noise ratio as well as envelope distortion. The results indicate a lack of overall support for the AAH with respect to vertical stratification of crickets in Kudremukh. However, a strong, independent effect of height of calling on signal degradation was found, with the ground being the worst layer for transmission and the mid-understorey (2 m) emerging to be good for all species with respect to all three measures of signal degradation. I then analysed the transmission data from a different point of view, exploring the possibility of evolution of signal structures that may confer some advantages in terms of signal transmission, given the habitat-induced constrains on signal propagation. The idea was to examine why certain species perform better than others in terms of signal transmission. This investigation was aimed at characterizing the effect of call features on signal attenuation. I found that temporal features of calls did not have a strong effect on height-specific signal attenuation. While call duration had no effect on signal attenuation, duty cycle did influence attenuation profiles of the calls, with high duty cycle calls performing better than low duty cycle calls. However, there was no interaction of height with the temporal features of calls, eliminating the possibility of these features being shaped by microhabitat or height dependent transmission characteristics. Spectral features of calls, on the other hand, affected signal attenuation very strongly. As expected, low frequency calls performed better than high frequency calls and pure tone calls fared much better than the broadband calls, especially on the ground and the canopy. To the best of my knowledge, this is the first study to carry out a rigorous quantification of microhabitat selection in Ensifera. This is also the first detailed examination of the Acoustic Adaptation Hypothesis at the microhabitat level, tested in a natural assemblage of coexisting invertebrate species.

No description available.