For years my core concern has been how to extract a quantitative metric (i.e. the number of bird calls) from TB of sound recordings obtained by placing autonomous sound recorders in natural areas where species of conservation interest are present. Ultimately, we would like to use some quantitative metric (like the number of calls recorded per time unit) as an index of population size to monitor populations over time.
This new package builds on previous R packages to read in .wav files, and provides two simple indices (in functions 'acoustic_diversity' and 'bioacoustic_index') based on standard biodiversity indices (e.g. Shannon diversity index) that reflect roughly how 'diverse' the acoustic energy is distributed across the recorded frequency spectrum. These indices do not identify certain patterns (i.e. species), and could be equally diverse in a very noisy urban environment as in a natural environment, so you have to be careful how you apply them.
However, the caveat of simplicity aside, the functions run relatively fast (calculating the index only takes a few min for a 2 hr recording) and in my preliminary screening for a bunch of sample files I had (where calls of a seabird species had been manually counted) the acoustic diversity was positively correlated with the number of calls. Critical settings are the min and max frequency (to specify the frequency range in which you would expect your target species, and filter out background noise) and the db_threshold, which effectively ignores faint background noise (but can of course also ignore faint target species vocalisations).
These functions will not replace all requirements for more sophisticated processing of sound recordings, but for some applications they may provide a very efficient approach to analyse a large number of recordings and obtain a quantitative metric that can be compared over time or space.