Image Source Method

I implemented the image source method for generating an impulse responses for a polygonal scenes based on the paper by Borish and tested it on three geometries. For each geometry I tested it with three source and listener pairs, generated the impulse response using my algorithm and then use python to convolve the impulse reponse with an audio signal.

For all of these test I assumed that each of the walls has an absorption coefficient of 30% and also that the sound sources were 1 meter is radius. Any listener inside the sound source was assumed to have zero attenuation.

The original audio signal was a wav file samples at 44,100 Hz, and can be found here.

Case 1: Small Rectangular Room

Dimensions: 8 x 5 x 3 (obj model)

Source: (4, 2.5, 1.5)
Listener: (4, 2.5, 1.5)
Impulse Response
Auralized Sound

Source: (1, 2.5, 1.5)
Listener: (7, 2.5, 1.5)
Impulse Response
Auralized Sound

Source: (3.7, 2.8, 1.2)
Listener: (7.8, 4.0, 2.7)
Impulse Response
Auralized Sound

Case 2: Large Rectangular Room

Dimensions: 32 x 20 x 12 (obj model)

Source: (4, 2.5, 1.5)
Listener: (4, 2.5, 1.5)
Impulse Response
Auralized Sound

Source: (1, 2.5, 1.5)
Listener: (7, 2.5, 1.5)
Impulse Response
Auralized Sound

Source: (3.7, 2.8, 1.2)
Listener: (7.8, 4.0, 2.7)
Impulse Response
Auralized Sound

Case 3: Gallery Room

Dimensions: 8 x 5 x 3 with an arched ceiling and no side walls (obj model)

Source: (4, 2.5, 1.5)
Listener: (4, 2.5, 1.5)
Impulse Response
Auralized Sound

Source: (1, 2.5, 1.5)
Listener: (7, 2.5, 1.5)
Impulse Response
Auralized Sound

Source: (3.7, 2.8, 1.2)
Listener: (7.8, 4.0, 2.7)
Impulse Response
Auralized Sound

The source code for the image source implementation can be found here, and the python script to convolve the audio can be found here (this requires Python 2.5, numpy, scipy and audiolab)