Authors:Lee, Keun-Sup; Abel, Jonathan S.; Välimäki, Vesa; Stilson, Timothy; Berners, David P.
Affiliation:CCRMA, Stanford University, Stanford, CA, USA; Dept. of Signal Processing and Acoustics, Aalto University, Aalto, Espoo, Finland; Universal Audio, Inc., Santa Cruz, CA, USA
High-quality artificial reverberators, which are well known, generally have large memory and computational costs. An alternative architecture attempts to manage quality separately for steady-state and transient signals. The proposed reverberator consists of an equalizing comb filter with an embedded convolution with a short noise sequence. To overcome the difficulty of unwanted periodicity with transients, the noise sequence is regularly updated. Several methods for updating the noise sequence, including a leaky integrator or a multiband structure are described. Informal listening tests are encouraging.
Download: PDF (HIGH Res) (1.5MB)
Download: PDF (LOW Res) (1.0MB)
Authors:Hasegawa, Tomomi; Tohyama, Mikio
Affiliation:Waseda University, Tokyo, Japan
It is difficult to analyze the sound from vibrating piano strings because the harmonics are not pure harmonics of the fundamental. Both time and frequency perspectives are required, especially since the initial transient produced by the hammer is a critically important part of the sound. A cluster line spectrum model (CLSM) appears to adequately capture the decay but a cluster time-sequence model (CTSM) is required to analyze the initial transient. The sound source characteristic of the piano string was revealed by using a direct estimate by CTSM.
Download: PDF (HIGH Res) (2.4MB)
Download: PDF (LOW Res) (1.4MB)
Authors:Wältermann, Marcel; Raake, Alexander; Möller, Sebastian
Affiliation:Deutsche Telekom Laboratories, Berlin, Germany
Traditional methods for evaluating degradations in a speech transmission system generally produce a single measure of quality without providing any diagnostic information about the nature of the degradation. In this research, a method is proposed to directly evaluate three orthogonal dimensions: discontinuity, noisiness, and coloration. Naïve listeners are able to distinguish the three dimensions after a brief training phase. The benefit to this approach is a reduced experimental effort that results from a limited number of scales and a reduced number of judgments per stimuli. The method was applied to two listening experiments.
Download: PDF (HIGH Res) (840KB)
Download: PDF (LOW Res) (275KB)
Authors:Koenig, Bruce E.; Lacey, Douglas S.
Affiliation:BEK TEK LLC, Clifton, VA, USA
When questioned as forensic evidence, audio recordings are frequently examined scientifically to insure that they have not been modified by being exported, edited, and then imported back to a recorder. For those recorders that use the Windows Media Audio (WMA) format, their authenticity can be determined, in part, by examination of the header data. Eleven audio recordings from three Olympus recorders were examined for changes in the digital header data, after being externally re-encoded by four common audio editing programs. All of the files were found to have undergone significant changes to the WMA header information by the re-encoding processes. Therefore, when conducting forensic authenticity analyses of WMA files, a detailed understanding of the header is required.
Download: PDF (HIGH Res) (1.1MB)
Download: PDF (LOW Res) (328KB)
Audio networking is coming of age, demonstrating a greater degree of interoperability between systems. Considerable progress has been made in standardization and common means are emerging for controlling audio devices connected to a network. IP-based data networking is set to replace telecoms-based networking as the method of choice for transferring media data. These topics and more were aired at the recent 44th International Conference, held in San Diego.
Download: PDF (1.7MB)