Menu:
WAFISEP
WaFiSep is the acronym of a project entitled "Rendering of 3D-Sound Scenes using a combination of Sound Source Separation methods and Wave-Field Synthesis". The project has been funded by the Ministry of Education and Science of the Spanish Government under the National Plan for R&D. The project started 1 Oct 2006 with a duration of 3 years.This project is part of a large coordinated project that comprises 4 sub-projects involving audio research groups of 5 Universities in Spain. The main project is call "Analysis, Classification and Synthesis for Sound Separation (AnClaS3)" and is coordinated by the group of the Technical University of Valencia.
Abstract
During the last years, Wave Field Synthesis (WFS) has become one of the most promising multi-channel sound reproduction systems, as demonstrate the increasing amount of scientific published papers. This system allows recreating spatial sound scenes with a high degree of immersion, and mainly, for a wide listening area. WFS is based on the wave theory concept of Huygens: all points on a wave front serve as individual point sources. A modification to the horizontal plane allows to use linear loudspeaker arrays fed by a corresponding driving signal calculated by means of signal processing algorithms, in order to accurately reproduce the acoustic field that would generate the virtual sources anywhere in the listening plane.
However, one of the main drawbacks of WFS is the necessity of having the original audio signal of each source to virtually locate it in the sound scene and thus synthesize the surround acoustic field. Most available recordings are stereo, so, no individual information of sources is available.
The first goal of this subproject consists on combining the sound source separation (SSS) techniques with WFS in order to synthesize, from standard recordings, very realistic spatial sound scenes. The SSS techniques show high complexity. There exist two main ways to implement it: the blind source separation (BSS) based on ICA and the separation based on models. In the first project stage, this problem will be tackled by means of the published separation algorithms of both types. To carry out with that, these algorithms will be adapted and implemented in order to be used in conjunction with WFS. The spatial interference between sources due to the non-ideal separation algorithm effects will be characterized and measured. Also, the improvement in the source allocation achieved by the WFS compared to the one obtained by the original stereo will be evaluated. In a posterior project stage, the new separation techniques developed by other groups in the joint project will be used, analyzed and compared.
A second group of objectives is focused on the improvement of the acoustic field synthesized by WFS systems, where the work will be carried on in three parallel research lines: 1) Room compensation methods for the disturbance due to the existing echoes inside the listening room. Multichannel inverse filter banks with sub-band processing placed before WFS algorithm will be used. 2) Synthesis of room acoustic responses by means of WFS compatible methods in order to increase the immersive feeling of these systems. New algorithms that combine in each particular case the advantages of the numeric methods FDTD (Finite-Differences Time Domain) and DWG (Digital Wave Guides) will be developed, and then extended to blocks for an efficient implementation. 3) Synthesis of sound sources with directivity features by means of the combination of beamforming and WFS techniques. Finally, a comprehensive subjective evaluation of the proposed techniques will be carried out, providing realistic data about their effectiveness.
The results of this project are directly applicable to render sound inside large enclosures, such as theatres and cinemas, and can bring to the public new and amazing immersive experiences. On the other hand, WFS presents meaningful applications to the engineering, i.e. architecture, where computational models of buildings can be reproduced with WFS thus allowing to walk virtually into the building before the construction.