Master of Arts / Department of Communication Studies, Theatre, and Dance / Timothy R. Steffensmeier / Geographical borders represent a clash of cultures. Those inhabiting or moving through borderlands struggle to maintain a sense of place and, in turn, an understanding of cultural collective memory. This project strives to understand how the visual and discursive elements that constitute the U.S.-Mexico border function rhetorically to communicate difference and establish place. By utilizing a social semiotics perspective, I analyzed visual rhetoric of the U.S.-Mexico border in the form of photographs and maps produced in both the United States and Mexico. Additionally, a theory of cultural memory was used to explore the confluence of events and rhetorical phenomena that shape the U.S.-Mexico border, and allow the U.S.-Mexico border to shape the rhetoric of the countries it divides. I argued that borders are inherently rhetorical and the intersection of visual elements, culture, place and memory make borders important to understand from an anthropological, and geographical perspective, as well as a rhetorical one. This project holds political and social implications for the relationship between the United States and Mexico, and reveals key findings regarding how cultural identity is negotiated in fragmented places like borderlands.

This paper studies the quality of synthetic speech produced by integrated circuit (IC) hardware using fixed-point arithmetic and Linear Predictive Coding (LPC). A theoretical model explaining the combined effects of finite wordlength and parametric model order is developed. This model is used to predict the results obtained in the experimental phase of this study. In the experimental phase, selected model utterances are synthesized under finite wordlength constraints using LPC parameters. The synthetic speech is evaluated in terms of the log area ratios which define objective speech quality as a parametric distance. A theoretical model is developed to predict the experimental results. Simulations of this model produce data that predict the experimental results. The same information is extracted from the model as that obtained from actually running the fixed-point synthesizer simulator. Since the predictions of the theoretical model agree quite well with the experimental measurements, it is concluded that fixed-point synthesizer performance can be predicted without actually running a complicated and expensive fixed-point synthesizer. Secondly, results obtained from either method clearly indicate that for 15 or 16 bits, ten is the best number of poles to use. Eight useable poles are indicated for 14 bits, while seven are indicated for 13 bits. Based on the results of this study, the use of less than 13 bits for fixed-point calculations is not recommended.

No description available.

A common recommendation for communication partners of people with hearing loss is to speak clearly, but how effective is this simple instruction? Does everyone produce clear speech using the same strategies? Is clear speech produced when given minimal instruction the same as that produced when competing background noise is present? The present study examined the acoustic characteristics of passage level speech produced in four different conditions. Twelve talkers (8 female, 4 male) with a mean age of 21 years were audio recorded reading three paragraph length passages. In the first condition talkers read each passage conversationally as though speaking to a friend. In the three experimental conditions, talkers were instructed to speak as clearly as they could, speak as clearly as they could in the presence of multitalker babble, and speak as clearly as they could in the presence of speech-shaped noise. The babble and noise were presented over headphones at a level of 75 dB SPL. Acoustic measures examined changes in rate, frequency, and intensity across condition. Results of this study help clarify what changes talkers make in response to instructions to speak clearly compared to conditions with competing background noise.

published_or_final_version / abstract / Speech and Hearing Sciences / Master / Master of Science in Audiology

No description available.

No description available.

No description available.

HMM-based systems for Automatic Speech Recognition typically model the acoustic features using mixtures of multivariate Gaussians. In this thesis, we consider the problem of learning a suitable covariance matrix for each Gaussian. A variety of schemes have been proposed for controlling the number of covariance parameters per Gaussian, and studies have shown that in general, the greater the number of parameters used in the models, the better the recognition performance. We therefore investigate systems with full covariance Gaussians. However, in this case, the obvious choice of parameters – given by the sample covariance matrix – leads to matrices that are poorly-conditioned, and do not generalise well to unseen test data. The problem is particularly acute when the amount of training data is limited. We propose two solutions to this problem: firstly, we impose the requirement that each matrix should take the form of a Gaussian graphical model, and introduce a method for learning the parameters and the model structure simultaneously. Secondly, we explain how an alternative estimator, the shrinkage estimator, is preferable to the standard maximum likelihood estimator, and derive formulae for the optimal shrinkage intensity within the context of a Gaussian mixture model. We show how this relates to the use of a diagonal covariance smoothing prior. We compare the effectiveness of these techniques to standard methods on a phone recognition task where the quantity of training data is artificially constrained. We then investigate the performance of the shrinkage estimator on a large-vocabulary conversational telephone speech recognition task. Discriminative training techniques can be used to compensate for the invalidity of the model correctness assumption underpinning maximum likelihood estimation. On the large-vocabulary task, we use discriminative training of the full covariance models and diagonal priors to yield improved recognition performance.

This study attempts to determine whether available ministerial speech training meets the current speech needs of active Protestant ministers by asking: 1. In the opinions of ministers, how comprehensive was their speech training? 2. In active ministers' opinions, what are the current speech needs of ministers? 3. According to the information obtained from school catalogs and the speech departments of the schools from which the ministers graduated, what speech training is available at these institutions? 4. According to a comparison of the speech needs (given by ministers) with the available speech training (indicated by the schools): A. Does the available speech training meet the current speech needs? B. If not, what recommendations are needed?