Lauri Savioja: Room Acoustics Modelling and Parallel Computation

When: November 6th, 2013 @ 5:00 PM – 6:00 PM

Where: Room 4.33, Informatics Forum, Crichton Street, University of Edinburgh

Seminar Title

Room Acoustics Modelling and Parallel Computation

Seminar Speaker(s)

Prof. Lauri Savioja (Aalto University, Finland)

Seminar Abstract

All room acoustic modeling techniques are computationally demanding due to the underlying complexity of sound propagation. Solving the wave equation numerically is the most rigorous way to model room acoustics, but at higher frequencies the computational costs become excessive. For this reason, at that range acoustic simulations are typically based on an approximation known as geometrical acoustics in which sound is supposed to act as rays. This presentation will give an overview of different room acoustic modeling techniques.

Traditionally, all the modeling methods have been implemented for the central processing unit (CPU) of the computer. However, recent trends in the development of graphics processing units (GPU) and other many-core architectures is resulting in a paradigm shift. Modern GPUs are massively parallel processors, whose peak computational performance can be 10X more than that of current CPUs. This presentation aims to give an overview of GPU architectures and discuss how to design algorithms that map well to a GPU.

The main focus of this presentation is to bind the two above mentioned areas together, and to present how GPUs can be utilized in room acoustic modeling. The case studies illustrate that both the numerical wave-equation solvers and geometrical acoustics techniques can gain remarkable performance boosts by GPUs. Current and future foreseen performance bottlenecks and challenges are discussed as well.

 

Speaker Bio

Lauri Savioja received the M.Sc., the Licentiate of Science, and the Doctor of Science in Technology, from the Helsinki University of Technology (TKK), Espoo, Finland, in 1991, 1995, and 1999, respectively. In all those degrees, he majored in computer science, and the topic of his doctoral thesis was room acoustic modeling.

He worked at the TKK Laboratory of Telecommunications Software and Multimedia as a researcher, lecturer, and professor from 1995 till the formation of the Aalto University where he currently works as a professor in the Department of Media Technology, School of Science. The academic year 2009-2010, he spent as a visiting researcher at NVIDIA Research.  His research interests include room acoustics, virtual reality, and parallel computation.

Prof. Savioja is a senior member of IEEE, member of ACM, and AES, and a life member of the Acoustical Society of Finland. From 2010 he has been an Associate Editor of the IEEE Transactions on Audio, Speech, and Language Processing.