Generating acoustic training data in R with sonicscrewdriver

Augmenting data is a common practice in machine learning. It is the process of creating new data from existing data (e.g. by adding noise). This is done to increase the size of the training data and to make the model more robust.

Recent work has created tools for performing data augmentation using the sonicscrewdriver package for R. A guide to using these tools can be found here: Generating acoustic training data in R with sonicscrewdriver, and on the package's website: Augmenting audio data in R with SonicScrewdriveR. The later will be updated to reflect any future changes to the package.

The updated package is already on CRAN: sonicscrewdriver R package.

All of the generateX() functions (generateNoise(), generateTimeMask(), generateTimeShift(), more to come) in sonicscrewdriver are designed to operate on Wave-like objects (Wave or WaveMC from tuneR of their Tagged equivalents) or a list of Wave-like objects. Similarly, all of these functions return a list of Wave-like objects. This means you can combine these functions to create complex data augmentation pipelines.

AudioMoth metadata in R

 The latest release (0.0.4) of SonicScrewdriveR is now on CRAN and includes two new functions for reading metadata from AudioMoth devices. 

audiomoth_config() parses the CONFIG.TXT written to the microSD card during device operation.

audiomoth_wave() parses the metadata written in the Comments field of the wave file metadata. This allows easy access to time, date, temperature (if supported by firmware) and other metadata as an R list.

Please raise GitHub requests with any issues (or even requests).

Bioacoustics and military history

My Bioacoustics and military history blog post on BioAcoustica.

Audubon Core Updates: Public Review of Proposals to Update Existing Standards

The Audubon Core Maintenance Group is advancing six proposals to update terms in the Audubon Core. A 30 day period for public comment is now open. It closes at the end of September. Three of the proposals create controlled vocabularies that were envisioned during the initial formulation of the Audubon Core, but were incomplete when the vocabulary was ratified. Three proposals are for adding new terms to Audubon Core that are important for biodiversity sound descriptions (i.e., metadata for biodiversity sound files).  To view the proposals, visit https://github.com/tdwg/ac/milestone/2 and view the individual proposal issues and their associated documents. To comment on the proposals, leave comments on the relevant proposal issue. If you are unable to create GitHub comments, send them to the Maintenance Group convener at steve.baskauf@vanderbilt.edu .

Generating a swept sine wave in R

This uses functionality from the sonicscrewdriveR package:

install.packages("devtools")

library(devtools)

install_github("edwbaker/sonicscrewdriver")

library(sonicscrewdriver)

Generate a Wave object with a sine sweep from 0 Hz to 1kHz over two seconds:

w <- sweptsine(f0=0,f1=1000, sweep.time=2, output="wave")

So good they named it once

Oecanthus pellucens (Scopoli, 1763) - the Italian tree cricket - was recently found breeding in Dungeness, Kent, UK (Sutton, Beckmann & Nelson, 2017).

There are a number of recordings of this species in the BioAcoustica repository: recordings of Oecanthus pelluscens on BioAcoustica.

Like many tree crickets it has a certain translucent quality, and it is this that Giovanni Antonio Scopoli alludes to when giving it the specific name pellucens. Torre-Bueno's A Glossary of Entomology gives the definition:

pellucid, pellucidity, pellucid, transparent, whether clear or coloured.

(This definition is somewhat incorrect, something transparent is clear, it can be coloured or colourless, but I digress)

 That Scopoli refers to this property is clear in the original description: "Caput album, subpellucens". Subpellucens is best translated as opaque.

All very well, but so far not particularly interesting. The (minor for most) interest comes from an alternative definition of pellucid, here borrowed form the Oxford English Dictionary.

Of music or other sound: clear and pure in tone.

Crickets in general are known for their pure bell like songs, and that of Oecanthus pelluscens is no exception. Shown below are three chirps from the recording above (time on x-axis, frequency on y-axis). This shows a strong resonance frequency, with three clear, regularly spaced, harmonics: a clear and relatively pure tone indeed.

Oecanthus pelluscens, a cricket so clear and pure it could be named just once.

Orthoptera Culture Group

Google Group for discussing the lab/zoo Culture of Orthoptera (grasshoppers/crickets/bush-crickets).