Simulating indoor/terrestrial sound propagation to test analysis workflows

Question

I'm currently trying to analyse multi-channel audio recorded in a cave (reverberant conditions). I'd like to check how well my bat calls are being localized, and the extent to which reverberance contributes to error in estimation of call duration.

I've tried simulating reflections by adding a synthetic signal over itself at various delays and amplitudes. This works to an extent, but is cumbersome and not always physically correct.

What methods/packages are out there to simulate more complex sound propagation?

Answer 1

There are two broad methods to simulate sound propagation that are very fast, and also make sense for most use cases: 1)image source model and 2)ray-casting. Both of these methods are inspired by methods to simulate light propagation in rooms. The approximations work when the wavelength of sound is smaller than the dimensions of the room. Most packages typically allow you to define the playback signal and location/s of source and mic arrays.

I'm a huge fan of the Python package pyroomacoustics and have used it to test the effect of reverb on acoustic localisation. The package allows you to simulate sound propagation in box-like rooms, and even load your own 3d mesh to simulate sound propagation in (e.g. scan of above-mentioned cave or room). Back when I was still using it, the ray casting was experimental - but the image source model implementation was stable.