Thesis (Ph. D.)--Architecture, Georgia Institute of Technology, 2006. / Brani Vidakovic, Committee Member ; Larry Kirkegaard, Committee Member ; Yves Berthelot, Committee Member ; Ruchi Choudhary, Committee Member ; Ning Xiang, Committee Member ; Godfried, Augenbroe, Committee Chair. Vita. Includes bibliographical references.

While humans can separate a sound of interest amidst a cacophony of contending sounds in an echoic environment, machine-based methods lag behind in solving this task. This thesis thus aims at improving performance of audio separation algorithms when they are informed i.e. have access to source location information. These locations are assumed to be known a priori in this work, for example by video processing. Initially, a multi-microphone array based method combined with binary time-frequency masking is proposed. A robust least squares frequency invariant data independent beamformer designed with the location information is utilized to estimate the sources. To further enhance the estimated sources, binary time-frequency masking based post-processing is used but cepstral domain smoothing is required to mitigate musical noise. To tackle the under-determined case and further improve separation performance at higher reverberation times, a two-microphone based method which is inspired by human auditory processing and generates soft time-frequency masks is described. In this approach interaural level difference, interaural phase difference and mixing vectors are probabilistically modeled in the time-frequency domain and the model parameters are learned through the expectation-maximization (EM) algorithm. A direction vector is estimated for each source, using the location information, which is used as the mean parameter of the mixing vector model. Soft time-frequency masks are used to reconstruct the sources. A spatial covariance model is then integrated into the probabilistic model framework that encodes the spatial characteristics of the enclosure and further improves the separation performance in challenging scenarios i.e. when sources are in close proximity and when the level of reverberation is high. Finally, new dereverberation based pre-processing is proposed based on the cascade of three dereverberation stages where each enhances the twomicrophone reverberant mixture. The dereverberation stages are based on amplitude spectral subtraction, where the late reverberation is estimated and suppressed. The combination of such dereverberation based pre-processing and use of soft mask separation yields the best separation performance. All methods are evaluated with real and synthetic mixtures formed for example from speech signals from the TIMIT database and measured room impulse responses.

This thesis describes the construction of an apparatus to measure reverberation time.

京都大学 / 0048 / 新制・論文博士 / 博士(工学) / 乙第13091号 / 論工博第4152号 / 新制||工||1675(附属図書館) / (主査)教授 髙橋 大弐, 教授 原田 和典, 教授 竹脇 出 / 学位規則第4条第2項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

In this thesis, a discussion on the effectiveness of acoustic design in public spaces is made. The auditory properties of a location have noteworthy implications on the success of a building's design and how a room is perceived. Depending on the requirements of each location, either a reverberant or sound-absorbing approach is best suited for the environment. Moreover, public health is negatively affected by long-term involuntary noise exposure. Because of this, there is an obvious demand for continued and expanded study in acoustic design. This thesis aims to challenge interior design choices made in four testing locations: a classroom, a musical practice room, an ambient performance space, and an office. Reverberation time is tested at each site using both a Digital Sound Level Meter application (Decibel X) and a 732A Digital Sound Level Meter paired with a series of external source sounds at set testing frequencies. Depending on the results of each trial, an evaluation of possible improvements to each location's aural properties is made.

A historical review is performed to study the impact of acoustical knowledge on concert hall developments. It shows that although acoustics developed from myth to real science, there is still a gap between its fast growing knowledge and relatively slow applications to improve designs. Architectural acoustics research and education shall help populating the tacit knowledge and experience of acousticians to reduce the gap between design and knowledge. The established paradigm in this field is to identify the performance goals of concert halls, recognize the available design information in different stages, and establish models to link them together. Placed in this general picture, this thesis focuses on providing design support for preliminary stage. It develops a model to link accessible design features with the most important acoustics performance index, reverberation time. A literature review on exiting reverberation time prediction methods shows that they are based on either too demanding or over-simplified for this stage. This study intends to develop a model that makes maximum use of available information and improves prediction accuracy in comparison with existing simplified methods. Through literature survey and data analysis, three factors (geometrical shape, non-uniform material distribution and scattering effect) are recognized as significant for reverberation time prediction. This thesis developed a simplified model taking these factors into consideration and calibrated this model with empirical data through Bayesian statistical method.

Abstract Noise has become an increasing public health problem and has become serious environment pollution in our daily life. This indicates that it is in time to control and reduce noise from traffic and installations in homes and houses. Today a plethora of products are available for business, but none for the private market. The project describes a start up of development of a sound absorbing wall panel for the private market. It will examine whether it is possible to make a wall panel that can lower the sound pressure level with 3 dB, or reach 0.3 s in reverberation time, in a normally furnished bedroom and still follow the demands of price and environmental awareness. To start the project a limitation was made to use the textiles available per meter within the range of IKEA. The test were made according to applicable standards and calculation of reverberation time and sound pressure level using Sabine’s formula and a formula for sound pressure equals sound effect. During the project, tests were made whether it was possible to achieve a sound classification C on a A-E grade scale according to ISO 11654, where A is the best, with only textiles or if a classic sound absorbing mineral wool had to be used. To reach a sound classification C, a weighted sound absorption coefficient (αw) of 0.6 as a minimum must be reached. The project resulted in that it is technical possible to achieve a sound classification C with only textiles even though for this project another combination was chosen for proceeding with the calculations, because of account to price and environmental awareness. The calculations showed that it is possible to lower the reverberation time to 0.3 s in a normally furnished bedroom with 7 wall panels, and to achieve a lowering of the sound pressure level with 3dB with 7 wall panels. This project showed promising results and leave openings for further research with only textiles and further calculations where more factors are taken under consideration to get more precise and reliable results. / Program: Textilingenjörsutbildningen

Thesis (Ph. D.)--University of Texas at Austin, 2006. / Vita. Includes bibliographical references.

The purpose of this report was to investigate the sound environments in lecture halls at Umeå University and to assess to whether the sound environments in these lecture halls are satisfactory and whether have good speech intelligibility or not. Based on a web survey, a number of lecture halls were selected for this investigation, namely Hörsal A, Hörsal C, N280, KB.F3.01, Ma121, KB.E3.01, N340 and N335. Assessing room acoustics includes many relevant parameters, therefore a delimitation has been made. Only installation noise, reverberation time and speech intelligibility have been investigated. The methods followed established Swedish standards i.e SS-EN ISO 16032:2004 for installation noise,” integrated impulse response method” SS-EN ISO 3382:2:2008/AC:2009 for reverberation time and direct FULL STI (speech transmission index) IEC 60268-16:2011 for speech intelligibility. An ocular inspection was carried out of all the selected lecture halls. The result from the STI test showed that all the lecture halls except for Hörsal A have a good speech intelligibility, if the audience has normal hearing and they are listening to a lecture in their native language. In Hörsal A there are three measuring positions that fails to achieve the threshold for good speech intelligibility. The conclusion for all of the measurements that has been made are that Ma121 and KB.E3.01 has a satisfactory sound environment and that Hörsal A, Hörsal C, N280, KB.F3.01, N340 and N335 need actions to improve their sound environment so they can fulfil the benchmarks in Swedish public health authority regulations in FoHMFS 2014:13 and/or SS 25268:2007.

This master’s thesis deals with the problems of interior acoustics. Within the scope of the thesis measurements were taken, followed by 3D modeling and subsequent design modifications of Olomouc acoustic music club Metro. The calculations used a specialized software (Odeon Room Acoustics Program). The result of the work is a proposal of suitable reconstruction and acoustic treatment, which will be achieved by optimalization of reverberation time, in an effort to improve the physical properties of the object.