Post provided by Abigail Searfoss and Nicole Creanza
Today, science extends beyond the research bench or the fieldsite more often than ever before. Scientists are continuously interacting with educators and the general public, and people are reciprocating the interest with a drive to be involved.
With this integration of science and the public, citizen-science efforts to crowdsource information have become increasingly popular (check out Zooniverse, SciStarter, NASA Citizen Science Projects, Project FeederWatch, and Foldit to get involved!). In the birding community, enthusiasts have been observing and recording birds for decades, but now there are methods for immediate data sharing among the community (eBird).
Two particular organisations have risen to the challenge of maintaining sites for collection and curation of these new recordings – The Macaulay Library at the Cornell Lab of Ornithology and Xeno-canto. These sites have developed an infrastructure to store hundreds of thousands of recordings, check recording quality and species identification, and disseminate the data. The main drawback to citizen-collected data is the variability in the collection methods and the quality of recordings. Imagine comparing a recording you made on your phone to one made with sophisticated professional-grade equipment. These two recordings would present very different challenges when extracting data.
So, for birdsong, it’s difficult to apply the same analysis procedure to each citizen-science recording. If researchers want to use this vast and valuable resource, they must take time-intensive measures to process the recordings. But, if citizen-science recordings of birdsong could be more easily analysed, they’d provide a geographically well-distributed source of information that isn’t easy to obtain as an individual researcher or small team.
Inception and Use of Chipper
In the Creanza Laboratory at Vanderbilt University, we’re particularly excited about citizen science recordings of birdsong. They provide a unique dataset of naturally occurring learned behaviour collected by enthusiastic birders across the globe. So we’ve developed methods to make it easier to use this wonderful resource.
When working on our latest project – uncovering geographic variation in chipping sparrow song – we decided to turn our in-house code for song analysis into an open-source, user-friendly desktop application to share with others. We designed the software because we saw a need for a method for analysing hundreds of recordings collected using a variety of devices in varying recording conditions. Hopefully it will increase the use of citizen-science recordings of birdsong in scientific research.
Ultimately, we named our software Chipper, reflecting both its first use in our chipping sparrow research project and its function, ‘chipping’ songs into its parts (syllables and notes). Chipper v1.0 can be downloaded for Mac, PC, and Linux from our GitHub repository; the source code can also be found here. You can find a full description of our methods in ‘Chipper: Open‐source software for semi‐automated segmentation and analysis of birdsong and other natural sounds’.
Chipper Streamlines Syllable Segmentation
Chipper streamlines your birdsong analysis process! You can load in a single song or a folder of songs to easily visualise, segment, and measure each one. By adjusting parameters you can remove unwanted signal and amplify the portion of the song you’re interested in for your analysis.
For instance, you can use sliders to change the high- and low-pass filters to remove unwanted low-frequency noise (e.g. urban noise) or high-frequency noise (e.g. crickets). You can normalise the amplitude to make sure you get the low amplitude portion of a song; often the beginning or ending of a song is much quieter as the bird is ramping up or down its volume.
The amount of signal retained can be adjusted so that only the top percent is kept. This feature is a huge help when it comes to removing unwanted calls or songs from other birds in the background that are at a much lower amplitude. As you adjust these parameters along with others in the dashboard, Chipper will automatically update the segmentation—the syllable markers will snap to the new signal transition.
Chipper was designed with easy-to-use features such as sliders to minimise the manual effort put into each song segmentation. This way, you don’t have to move a cursor to every x- or y-position to measure syllable durations and frequencies. Instead, the visualisation of the syllable segmentation will dynamically update based on the user-adjusted parameters, providing markers of syllable beginnings and endings. These help you to clearly visualise how you’re breaking the song into its constitutive parts—syllables, notes, and silences. Chipper then uses these markers to calculate and provide the user with raw measures of the frequencies and durations of syllables.
Choosing the thresholds for noise and syllable similarity
You can get more than just frequency and duration measures from Chipper! All you need to do is adjust two more parameters – noise threshold and syllable similarity threshold. These will allow Chipper to provide you with more accurate note measurements and a representation of the song syntax. As these two parameters are difficult to determine for a set of songs, we’ve created two more user-friendly dashboards to visualise your parameter choice, ultimately helping you identify the best threshold for your data.
This video explains how to download and use Chipper’s primary feature – syllable segmentation. You can also find a step-by-step guide to using Chipper here.
Contribute to Chipper!
If you run into any trouble downloading and running Chipper, find any bugs, or would like to suggest changes or improvements to Chipper, please open an issue on GitHub.
We created Chipper using open-source software so that the community can contribute to improving and adding new functionality to it. An easy place to start would be to adjust or add measurements to the output from Chipper’s analysis. You can do this by editing the analysis.py script. All information from segmentation and threshold determination have been added as attributes of the Song class and can be used for additional calculations. If you’d like to contribute your changes to Chipper so others can also benefit, please submit a pull request on GitHub.
Let us know what you think!
While Chipper was designed primarily for analysing birdsong, any WAV file can be loaded into Chipper for assessment. We’re excited to see the other areas of research where Chipper will prove to be useful! Be sure to tweet us @CreanzaLab with any research enabled by the use of Chipper!
If you want to learn more, you can see our recent paper (Searfoss et al., Animal Behaviour 2020) where we use Chipper to find geographic patterns in chipping sparrow song!
For more information on Chipper, read our Methods in Ecology and Evolution article ‘Chipper: Open‐source software for semi‐automated segmentation and analysis of birdsong and other natural sounds’. No Subscription Required.