This is a Matlab implementation of the room acoustic simulator "Scattering Delay Network" (SDN) as described in [1] and [2]. If you use this code in your research publication, please make sure to cite [1].
Please notice that this Matlab implementation was written quickly in 2010 for the purpose of showing results in [2], and is extremely slow, possibly because of the sample-by-sample operation and Matlab's inefficient handling of object-oriented programming. The algorithm itself is actually orders of magnitude faster than even fft-based convolution. A dynamic C++ implementation exists which uses << 1% of a single core on a modern CPU.
To get acquainted with the code, you can look into the script 'sdn_examples.m'. You can also check that everything is working as expected by running 'sdn_tests.m'.
Please notice that the SDN algorithm is protected by USPTO patent 8,908,875 [3]. More specifically, the part of package that implements the patent is contained in the file "Simulation.m". If you'd like to use this software for any reason other than non-commercial research purposes, please contact enzodesena AT gmail DOT com
You need the Signal Processing Toolbox.
No need to install: just clone/download and you are ready to go!
% Example script generating the SDN output for:
% - a 5x5x5 m room,
% - a source positioned at (0.3,0.5,0.9)
% - an omnidirectional microphone positioned at (0.4,0.1,0.4)
% - wall reflection coefficient of 0.9 followed by a low-pass butterworth
% filter
% - sampling frequency: 44100
% - calculating 10000 samples (i.e. ~0.23 sec)
room = Room();
room.shape = Cuboid(5,5,5);
d=fdesign.lowpass('N,F3dB',5,15000,44100);
for i=1:6
room.wallAttenuations{i} = 0.9;
for j=1:5
room.wallFilters{i}{j} = design(d,'butter'); %dfilt.delay(0);%
end
end
source = Source();
source.position = Position(0.3,0.5,0.9);
source.signal = Signal([1, zeros(1,9999)], 44100);
microphone = Microphone();
microphone.position = Position(0.4,0.1,0.4);
sim = Simulation();
sim.room = room;
sim.source = source;
sim.microphone = microphone;
sim.NSamples = 10000;
sim.frameLength = 1;
output = sim.run();
plot(output)Run sdn_tests.
The conceptual work to develop SDN was carried out by all authors in [1] and [2].
The Matlab software in this repository was written by Enzo De Sena while he was a PhD student at King's college London in 2010.
- Enzo De Sena - desena.org
[1] E. De Sena, H. Hacıhabiboğlu, Z. Cvetković, and J. O. Smith III "Efficient Synthesis of Room Acoustics via Scattering Delay Networks," IEEE/ACM Trans. Audio, Speech and Language Process., vol. 23, no. 9, pp 1478 - 1492, Sept. 2015.
[2] E. De Sena, H. Hacıhabiboğlu, and Z. Cvetković, "Scattering Delay Network: An Interactive Reverberator for Computer Games," in Proc. 41st AES International Conference: Audio for Games, London, UK, February 2011.
[3] E. De Sena, H. Hacıhabiboğlu, and Z. Cvetković, "Electronic Device with Digital Reverberator and Method", US Patent n. 8,908,875, filed 2/2/2012, granted 09/12/2014.
This project is licensed under the AGPL License - see the LICENSE file for details. Also notice that parts of the code are protected under USPTO patent n. 8,976,977.