Catharina Karlsson is currently using machine learning and automated bio-acoustic systems to study frogs in Borneo. She is really knowledgeable and likely willing to share her resources on how to assemble and implement the systems. I would talk to her.

As far as your question of "How effective" it would be. I would say that it would be very effective given the proper sample design. It all depends on how well you deploy your acoustic equipment.

Russell Gray I think she is asking more if machine learning is feasible? My immediate answer is yes. We work on multi-species assemblages, we might have as much as 6-7 species calling within just a few seconds, it challenges the algorithms a a bit. Especially as the data is from omni-directional microphones so there's a lot of background noise. Not impossible though, just need to ensure the training is done thoroughly. Have a look at the papers by Bedoya 2014 (Automatic Recognition of Anuran Species based on Syllable identification). We are also working on implementing deep learning on our data - they've done it for bats in our lab (I think the paper is in review) and there was just a paper released on using it on birds I think? (I can't find it or remember the author off the top of my head, might have been in Methods in Ecology). Birds are in my experience a pretty good model species though as their calls are quite complex so easy for the algorithms to identify, we have more issues with the frogs as the call itself is often quite "simple" in it's pattern (then put together into a more complex call series), I don't know how primate calls look like in spectrograms though.

The Erin M. Bayne Lab (https://www.researchgate.net/profile/Erin_Bayne/projects ) at the university of alberta has a big group that have been working for several years on bioacustic and automatic detection of birds in the boreal forest. Some of their work might be useful to you.