The Multi-dimensional Digital Waveguide Mesh: applications in modelling of room acoustics and of the vocal tract
Damian Murphy, Media Engineering Group, Dept of Electronics, University of York
The Multi-dimensional Digital Waveguide Mesh (DWM) was first introduced as a modeling technique for music acoustics applications in 1993 and is an extension to the 1D digital waveguide used extensively for physical modelling sound synthesis. The DWM has been shown to be appropriate for modelling 2D structures such as plates and membranes and in 1994 was first applied to problems in the field of room acoustics simulation. Work has been steadily growing in this area across a wide range of international research groups, with York (UK) focusing efforts on making improvements to the technique for room acoustics modelling applications. Notable results include the use of 2D triangular mesh topologies for improved reverberation simulation, modelling 3D enclosed spaces with a tetrahedral mesh topology, developing an improved anechoic DWM boundary and analysing the spatial properties of DWM-based room impulse response measurements. Current work is concentrated in three main areas and this seminar presents recent results from these parallel studies. RoomWeaver is a DWM Integrated Development Environment that allows ease of development and application of DWM based models for virtual acoustic spaces.
RoomWeaver introduces the use of hybrid mesh-types incorporating wave-based and finite difference-based DWM elements through the use of 2D and 3D KW-pipes. The KW-pipe allows the geometrically flexible boundaries of wave-based models with the computational efficiency of the finite difference approach showing significant improvements in terms of execution time and memory use. Secondly, new boundary mesh-types are being investigated with recent work implementing a diffusive reflection layer, offering variable user control of how specular or diffuse a surface may be. The third research area focuses on Digitract, a physical modelling speech synthesis research tool based that extends the established 1D digital waveguide vocal tract simulation through the use of the 2D DWM. The 2D vocal tract model can be seen to offer improved control over formant bandwidths and improved naturalness in the production of speech sounds. Future work will look to incorporate Digitract as part of RoomWeaver, giving more flexible definition and control of speech synthesis parameters.