AES Vienna 2007
Home Visitors Exhibitors Press Students Authors
Visitors
Technical Program
Paper Sessions

Spotlight on Broadcasting

Spotlight on Live Sound

Spotlight on Archiving

Detailed Calendar

Convention Planner

Paper Sessions

Workshops

Tutorials

Exhibitor Seminars

Application Seminars

Special Events

Student Program

Technical Tours

Technical Council

Standards Committee

Heyser Lecture









Last Updated: 20070405, mei

P23 - Analysis and Synthesis of Sound

Tuesday, May 8, 09:30 — 12:00

Chair: Gerhard Graber, Technical University of Graz - Graz, Austria

P23-1 Sound-Transformation and Remixing in Real-TimeHannes Raffaseder, St. Pölten University of Applied Sciences - St. Pölten, Austria
Starting with a short overview on some very basic principles of sound perception, this paper acts on the assumption that recording, storage, editing, and reproduction of audio signals have compensated for at least some of these principles and, therefore, have significantly changed human listening habits. Reflecting on these changes, the idea of sound transformation and remixing in real-time is suggested as part of the performance and composition process. Some techniques are explored and a number of implementations are introduced.
Convention Paper 7132 (Purchase now)

P23-2 Hybrid Time-Scale Modification of AudioPatrick-André Savard, Philippe Gournay, Roch Lefebvre, Université de Sherbrooke - Sherbrooke, Quebec, Canada
This paper presents a novel technique for time-scale modification (TSM), which integrates time-domain and frequency-domain processing. The method relies on frame-by-frame classification to choose between different techniques adapted to different signal types. Provisions are taken to seamlessly switch between techniques. The result is a more universal TSM algorithm that yields continuous high quality results on a wider range of audio signals. The method is tested on mixed-content signals and formal listening tests results are discussed.
Convention Paper 7133 (Purchase now)

P23-3 New Audio Editor Functionality Using Harmonic SinusoidsWen Xue, Mark Sandler, Queen Mary, University of London - London, UK
This paper introduces the application of harmonic sinusoid model in an audio editor. The harmonic sinusoid model is a parametric model for representing pitched audio events, allowing amplitude and pitch evolution. While standard audio editors enable the user to select a time or frequency range to edit, with the harmonic sinusoidal parameters estimated in phase, we are able to select a pitched event and edit it as if it were separated from the background. The user interface is designed as a simple one-click selection, while the user is given further options for better results.
Convention Paper 7134 (Purchase now)

P23-4 Measurement and Optimization of Acoustic Feedback of Control Elements in CarsAlexander Treiber, Gerhard Gruhler, Heilbronn University - Heilbronn, Germany
Acoustical quality of control elements in cars is increasingly important for manufacturers in order to improve the quality, appearance, and security of their products. This paper presents methods and tools used in an ongoing research project. The project’s goal is to support the industry with the definition of suitable parameters and limits as well as to develop realizable proposals for measuring equipment. Jury tests are hereby used to create the scientific basis for the hearing-related benchmarking of signals.
Convention Paper 7135 (Purchase now)

P23-5 On the Training of Multilayer Perceptrons for Speech/Nonspeech Classification in Hearing AidsLorena Álvarez, Enrique Alexandre, Lucas Cuadra, Manuel Rosa-Zurera, Universidad de Alcalá - Alcalá de Henares, Spain
This paper explores the application of multilayer perceptrons (MLP) to the problem of speech/nonspeech classification in digital hearing aids. When properly designed and trained, MLPs are able to generate an arbitrary classification frontier with a relatively low computational complexity. The paper will focus on studying the key influence of the training process on the performance of the system. An appropriate election of the training algorithm will help to provide better classification with a lower number of neurons in the network, which leads to a lower computational complexity. The results obtained will be compared with those obtained from two reference algorithms (the Fisher linear discriminant and the k-Nearest Neighbour), along with some comments regarding the computational complexity.
Convention Paper 7136 (Purchase now)