I work with 4-22 mics on a regular basis in the field. A fixed array frame can't always be setup in the field site (a cave, and sometimes narrow spaces).
I would like to have 'free-form' arrays whose geometry can be estimated post-hoc without any/too many measurements. This seems possible, and is indeed done with camera arrays (where a common calibration object is moved around the scene).
Does anyone know if such techniques exist or have used them?
You might be relieved to hear that there are a small set of such algorithms out there! There is a nice review of the various types of workflows available in Plinge et al. 2016 - where they discuss the different types of scenarios (un/synchronised playback and recording devices, playback types etc.).
I was involved in a collaboration based on the Structure-From-Sound (SFS) approach proposed by Zhayida et al. 2016. SFS provides a flexible workflow that needs a playback from a common source, with no other restrictions. The time-difference-of-arrivals are used to infer microphone positions. More recently I was part of a collaboration the positions of microphones in a cave were inferred using ultrasound playbacks (paper link).
There is a caveat, to my knowledge, there are no open-source, user-friendly packages available yet, though do I hopewe will end up seeing one such package out in the near future.
