The sustained increase in computing performance over the last decades has brought enough computing power to perform significant audio processing in affordable personal computers. Following this revolution, we have witnessed a series of improvements in sound transformation techniques and the introduction of numerous digital audio effects to modify effectively the time, pitch, and loudness dimensions of audio signals. Due to the complex and multi-dimensional nature of timbre however, it is significantly more difficult to achieve meaningful and convincing qualitative transformations. The tools currently available for timbre modifications (e.g. equalizers) do not operate along perceptually meaningful axes of singing voice timbre (e.g. breathiness, roughness, etc.) resulting in a transformation control problem. One of the goals of this work is to examine more intuitive procedures to achieve high-fidelity qualitative transformations explicitly controlling certain dimensions of singing voice timbre. Quantitative measurements (i.e. voice timbre descriptors) are introduced and used as high-level controls in an adaptive processing system dependent on the characteristics observed in the input signal. / The transformation methods use a harmonic plus noise representation from which voice timbre descriptors are derived. This higher-level representation, closer to our perception of voice timbre, offers more intuitive controls over timbre transformations. The topics of parametric voice modeling and timbre descriptor computation are first introduced, followed by a study of the acoustical impacts of voice breathiness variations. A timbre transformation system operating specifically on the singing voice quality is then introduced with accompanying software implementations, including an example digital audio effect for the control and modification of the breathiness quality on normal voices.

This thesis examines the factors that affect categorical perception of Mandarin tones by English, Thai and Mandarin speakers. A pre-study and a main study were conducted. The pre-study was carried out to investigate the correspondence between Mandarin and Thai tones. The main study compared the categorization of two Mandarin tone contrasts, one between the high level tone (T1) and falling tone (T4), and the other between the rising tone (T2) and falling-rising tone (T3). The results suggest that linguistic experience, natural auditory sensitivity and perceptual training all affect Mandarin tone categorization. among which, the effect of linguistic experience is most noticeable. This study also demonstrates that Best et al.'s (1988, 1994, 1995) Perceptual Assimilation Model (PAM) is useful in characterizing not only segmental assimilation patterns, as illustrated in many previous studies, but also suprasegmental categorization patterns.

Improving the performance of the wireless multihop ad hoc networks faces several challenges. In omni-directional antenna based solutions, the use of the RTS/CTS mechanism does not completely eliminate the hidden-terminal and exposed-terminal problems. Deafness is an additional challenge to the directional antenna based solutions. This dissertation, first develops analytical models for quantifying the throughput and delay in wireless multihop ad hoc networks. The models consider the impact of hidden terminals using the realistic signal to interference and noise ratio model and consider random node distribution. The proposed analysis is applicable to many wireless MAC protocols and applications. The analytical results reveal several important issues. The first issue is quantifying the impact of adjusting the transmission range on the throughput and delay in wireless multihop ad hoc networks. The other issue is the hidden terminal region is closely related to the distance between the transmitter and the receiver. Thus, it is possible to adjust the transmission range to optimize the whole network performance. These results provide important guidelines for network planning and protocol optimization in wireless multihop ad hoc networks. Second, it proposes a new Enhanced Busy-tone Multiple Access (EBTMA) medium access control (MAC) protocol for minimizing the negative impact of both the hidden-terminal and the exposed-terminal problems. The new protocol can also enhance the reliability of packet broadcasts and multicasts which are important for many network control functions such as routing. Different from other busy-tone assisted MAC protocols, the protocol uses a non-interfering busy-tone signal in a short period of time, in order to notify all hidden terminals without blocking a large number of nodes for a long time. In addition, the proposed EBTMA protocol can co-exist with the existing 802.11 MAC protocol, so it can be incrementally deployed. Third, it investigates how to support the directional antennas in ad hoc multihop networks for achieving higher spatial multiplexing gain and thus higher network throughput. A new MAC protocol called Dual Sensing Directional MAC (DSDMAC) protocol for wireless ad hoc networks with directional antennas is proposed. The proposed protocol differs from the existing protocols by relying on a dual sensing strategy to identify deafness, resolve the hidden-terminal problem and to avoid unnecessary blocking. Finally, this dissertation provides important results that help for network planning and protocol optimization in wireless multihop ad hoc networks in quantifying the impact of transmission range on the throughput and the delay. The accuracy of these results has been verified with extensive discrete event simulations. / Graduate

Timbre is a multidimensional attribute of sound and depends to a large extent on its spectral content. The evaluation and control of timbre is a task commonly performed by sound engineers, loudspeaker designers, and subjects participating in listening tests on the quality of transmitted and reproduced sound. Such listening tasks require specific listening abilities. / This dissertation presents a training method that aims at developing memory for timbre, sensitivity to timbre changes, and listening strategies involving disciplined auditory attention and efficiency. The physical timbre space is divided into categories defined by the center frequency of standard octave and third-octave resonances. This simplification of the physical timbre space allows the memorization of a limited set of perceptual timbre categories or references that can be used to evaluate other timbres. / The proposed method combines the use of computer software for the presentation and evaluation of exercises and individual tutoring sessions with an instructor. The software monitors the actions of listeners during equalization adjustment tasks and allows data to be collected on the sequence of problem solving operators that are used. Performance indicators can thus be compiled and used for the assessment of skill level achieved by the listeners. A global performance index is proposed. / A listening test involving the adjustment of multiple peaks and dips in a sound's spectrum is presented, in which the performance of a group of experienced professionals involved in sound recording and audio post-production was compared with a group of student subjects with comparatively little listening experience but who were trained with the method described in the present work for a period of 6 months. Results indicate that overall the experimental student group outperformed the control group of professionals. Suggestions for further improvement of the training method and performance level assessment are proposed.

Characterization of nonlinear systems is a challenging task as the output can not be expressed simply in terms of input signal. Therefore, a universal analysis method is essential to simplify this procedure. Modeling of the input signal is a crucial part of such analysis. In this thesis, multi-tone representation is employed to model arbitrary, stochastically not well-defined input signals and thereafter characterize nonlinear systems. In order to verify the validity of multitone representation, multi-tone modeling concept is primarily applied to real life amplifier characterization in single amplifier configuration. This experiment demonstrated potential of multi-tone modeling concept in nonlinear system characterization and encouraged application of the concept to analysis of feedforward linearizers, which are complicated systems due to the presence of two nonlinear amplifiers and the requirement of strict amplitude, phase and delay matching within two loops of the circuit. It has been assumed that main and error amplifiers can be modeled with third order AM/AM nonlinearities and there exists no delay mismatch within the loops. Hence, closed form expressions relating the main and adjacent channel power at the output of the feedforward system to the system parameters are obtained. The developed model is verified by RF and system simulations. As a result, a mathematical handy tool to specify circuit parameters rapidly for optimum linearity performance and efficiency is achieved.

No description available.

With the volume of digital media available today, automatic music recommendation services have proven a useful tool for consumers, allowing them to better discover new and enjoyable music. Typically, this technology is based on collaborative filtering techniques, employing human-generated metadata to base recommendations. Recently, work in content-based recommendation systems have emerged in which the audio signal itself is analyzed for relevant musical information from which models are built that attempt to mimic human similarity judgments. The current state-of-the-art for content-based music recommendation uses a timbre model based on MFCCs calculated on short segments of tracks. These feature vectors are then modeled using GMMs (Gaussian mixture models). GMM modeling of frame-based MFCCs has been shown to perform fairly well on timbre similarity tasks. However, a common problem is that of hubs , in which a relative small number of songs falsely appear similar to many other songs, significantly decreasing the accuracy of similarity recommendations. In this thesis, we explore the origins of hubs in timbre-based modeling and propose several remedies. Specifically, we find that a process of model homogenization, in which certain components of a mixture model are systematically removed, improves performance as measured against several ground-truth similarity metrics. Extending the work of Aucouturier, we introduce several new methods of homogenization. On a subset of the uspop data set, model homogenization improves artist R-precision by a maximum of 3.5% and agreement to user collection co-occurrence data by 7.4%. We also find differences in the effectiveness of the various homogenization methods for hub reduction, with the proposed methods providing the best results. Further, we extend the modeling of frame-based MFCC features by using a kernel density estimation approach to non-parametric modeling. We find that such an approach significantly reduces the number of hubs (by 2.6% of the dataset) while improving agreement to ground-truth by 5% and slightly improving artist R-precision as compared with the standard parametric model. Finally, to test whether these principles hold for all musical data, we introduce an entirely new data set consisting of Indian classical music. We find that our results generalize here as well, suggesting that hubness is a general feature of timbre-based similarity music modeling and that the techniques presented to improve this modeling are effective for diverse types of music.

Efter operation och anestesi får patienter ofta en negativ påverkan på magsäck och tarmar. Illamående och kräkningar är ett stort problem och många har svårt att komma igång med intag av föda och normal tarmfunktion då magsäcken och tarmarna ”står stilla”. Flera faktorer bidrar- bl.a. smärtan, det kirurgiska traumat och de läkemedel vi ger i samband med anestesin. Av de senare är opioider, d.v.s morfin och morfinliknande läkemedel, starkt bidragande. I detta avhandlings- arbete har opioiders effekter på magsäckens motilitet studerats. Med ett absorptionstest (paracetamolmetoden) studerades hos frivilliga hur opioiden remifentanil påverkar magsäckstömning och om kroppspositionen har betydelse för tömningshastigheten ut i tarmen. Remifentanil fördröjde magsäcks-tömningen och under pågående opioid behandling hade kroppspositionen ingen större betydelse, vilket det däremot hade under kontrollförsöken. Med samma metod jämförde vi hos patienter två anestesimetoder och studerade magsäcks-tömning direkt efter en operation. Ingen skillnad kunde påvisas mellan en opioidbaserad och en opioidfri anestesi, men inom respektive grupp var det en stor variation i magsäckstömning mellan individerna. Med en barostat studerades tonus i övre delen av magsäcken. Hos hälften av de frivilliga orsakade remifentanil en ökning av tonus och hos den andra hälften en minskning av tonus. Vidare undersöktes hos en grupp patienter opioiden fentanyls påverkan på den elektriska aktiviteten i magsäcken. Med en elekroga-strograf (EGG) registrerades de långsamma elektriska vågor som koordinerar muskelrörelserna i magsäcken. Hos hälften av de undersökta påverkades aktiviteten av fentanyl med en sänkt vågfrekvens eller upphörande av vågor, medan aktiviteten var opåverkad hos den övriga hälften. För att finna en förklaring till variationen gjordes genetiska analyser av genen för opioidreceptorn hos de undersökta i barostat och EGG studierna. Variationer i genomet, s.k. polymorfism, var inte associerad till utfallen i studierna. Studierna har visat på att opioider har en uttalad effekt på magsäckens motilitet och att den varierar kraftigt mellan individer. Polymorfism i genen för opioid- receptorn förklarade inte skillnaden mellan individer. Direkt efter operation bidrar sannolikt andra faktorer än anestesimetod till det variabla utfallet i magsäckstömning. / After anesthesia and/or surgical procedures, gastrointestinal motility is commonly impaired. The causes are multifactorial, with surgical trauma, pain and perioperative drugs playing a major role. This thesis explores opioid effects on gastric motility in healthy volunteers and patients undergoing surgery. Gastric emptying was studied by an absorption test (paracetamol method), and in healthy volunteers a remifentanil infusion delayed gastric emptying. Body position altered emptying during the control situations, but not during the remifentanil infusion. Further, two anesthetic methods were compared and no differences were found in immediate postoperative gastric emptying between a remifentanil/propofol based intravenous anesthesia and an opioid free inhalational anesthesia, although the interindividual variability was high. Proximal gastric tone was studied using a gastric barostat. An infusion of remifentanil caused two patterns of reaction regarding gastric tone, with half of the subjects increasing and half decreasing in gastric tone. Gastric myoelectrical activity was evaluated with electrogastrography (EGG), and a bolus dose of fentanyl caused a decrease in frequency of the gastric slow waves or disrupted this activity. However, the activity was unaffected in half of the investigated subjects. Analysis of polymorphisms (A118G and G691C) in the µ-opioid receptor gene was performed to find an explanation for the great interindividual variations seen in the barostat and EGG studies, but no association could be found. These studies have shown that opioids have pronounced effects on gastric motility with variable individual responses that are difficult to predict. Polymorphisms in the µ-opioid receptor gene could not explain the variations. Postoperatively, other factors might contribute more than opioids to the impairment in gastric motility. / ISSN 1652-4063

The limiter is a commonly used device in communication receiving systems to remove the amplitude variations of the received signal, but it is usually observed that limiter degrades the envelope detection performance of a single tone. In this thesis, it is demonstrated that the limiter-generated third harmonic can be used to improve the envelope detection performance of a single tone over that of the linearly processed fundamental. Differences in the probability distributions of the limiter-generated harmonics cause differences in their detection probabilities, which lead to differences in the performance of subsequent envelope detection. Comparison of the envelope detection performance of the limiter-generated third harmonic and the input to the limiter shows a maximum detection probability gain of 1.12 and also error probability gain of 2.4 in linear scale, whereas the fundamental harmonic does not show any practically significant gain. The envelope detection performance of the vector sum of the limiter-generated fundamental and third harmonics is also evaluated. This combination provides better performance than do the individual harmonics, especially for a large clipping parameter of the limiter. The combined harmonics shows a maximum detetction probability gain of 1.15 and error probability gain of 14, over that of the envelope detection performance of a tone without limiter. It is also observed that the envelope detection performance of a tone with limiter-generated higher harmonics shows non-monotonic behaviour as functions of either noise or the limiter-clipping parameter, which is a signature of stochastic resonance. The theoretical results from earlier research on the envelope detection of a single tone embedded in additive white Gaussian noise, both with and without a limiter, are presented and shown to match our simulation results. In previous work when a limiter was used in the envelope detection of a single tone, only the envelope of the fundamental harmonic was considered under very specific conditions. By contrast we also take notice of the limiter-generated higher harmonics and obtain improved envelope detection performance in the detection of a single tone.

Thesis (Ed.D.)--Teachers College, Columbia University, 1996. / Issued also on microfilm. Includes tables. Sponsor: Lenore M. Pogonowski. Dissertation Committee: Harold F. Abeles. Includes bibliographical references (leaves 123-125).