Monongahela

A site dedicated to the review and analysis of potential sasquatch vocalizations, Sasquatch Bioacoustic combines techniques from the domains of intelligence collection, audio analysis and bioacoustic studies to examine the evidence of sasquatch through their vocalizations. ~Monongahela

Friday, December 23, 2011

Using Audacity Spectrograms to Review Audio - A Cheat Sheet

This is a quick tutorial and list of actions that will get the first time user up and running with Audacity, for the purpose of reviewing their audio files visually, with spectrograms. This assumes you are already familiar with how to open an audio file in Audacity.

-Start Audacity and maximize the application's window to fill your whole screen
-The first time you use this tutorial you'll want to type "Ctrl-P" to open the Audacity Preferences window
-In the Audacity Preferences window select "Spectrograms" to bring up the associated options
-In the spectrogram options set window size to "4096 - most narrowband" (later versions of Audacity allow for much higher settings, but greater than 8192 may be of little use for this purpose)
-Set Maximum Frequency to 1200, and click the box to "Show the spectrogram using grayscale colors" (Gray scale is easier to work with until you're more familiar with spectrograms)
-Select OK to save your settings

See the image below for added examples of these settings...



-Now select File>Open and choose an audio file to load into Audacity
-Grab the bottom edge of the sound track when it opens and drag it down to fill in the empty space in the main application window (or you can hold down the Ctrl and Shift keys simultaneously and hit the "F" key to achieve the same effect)

-At the top left of the sound track, the file's name will appear with a small down arrow to the right of the name
-Select that arrow and a drop down menu provides options to apply to the track, select "Spectrum" or "Spectrogram" in newer versions of Audacity


-In the resulting spectrogram view, use your mouse to select a segment of audio to examine.
-Start with 10-20 seconds until you've become comfortable with the types of patterns to look for.
-Select the button in the top right of the main toolbar to "Fit Selection", this will cause the view to zoom in to the audio you've selected with your mouse


-Use the scroll bar at the bottom of the audio file window to jump the audio right one segment of time (based on the duration you selected above).

-Scroll through the audio a bit at a time and watch for dark spots in the spectrogram which indicate an audible sound has been recorded
-When you spot something worth listening to, click your mouse just to the left of the dark spot and then hit the space bar on your keyboard to begin playback
-Listen to the sound, and after you've heard it, hit the space bar again to stop playback

Here's an example of how coyote howls appear in an audacity spectrogram set up as discussed above


-Use this method to scroll, view, and selectively play back audio until you've completed the entire file
-If you find something worth isolating, use your mouse to select that section of audio and then the menu options File>Export Selecttion to save that segment to a new file.
-One of the biggest advantages spectrogram review has over the normal "waveform" review method (where you watch for big spikes in the audio file's waveform view), is how spectrograms show you howls that are otherwise hidden in the waveform. As an example compare the spectrogram view of the coyote howls above, to the waveform view of the same coyote howls, below. In waveform they are complete invisible and would go unheard if you weren't lucky enough to play back that segment of audio.

 
-After you've become comfortable scrolling through spectrograms, recognizing interesting patterns (sounds), and playing them back, you should try widening the view on your screen
-A three minute segment (vice the 20 seconds described above), gives good visual detail when in full screen mode, and allows you to review hours of audio very quickly (as compared to listening to the entire file)
-This is what makes it possible to record 9 hours of audio every night, and then quickly review it in under thirty minutes at some future date (more recording means more chances of catching a sasquatch vocal)

Other special notes:
-Most good sasquatch vocals occur between 500 and 800 Hertz, but can appear above or below that (although rarely seen above 1200 Hertz)
-Howls will look like long flat or wavering horizontal lines (-----), whoops look like the forward slash on your keyboard (/), woodknocks look like tall vertical lines (|)
-Dogs most often appear in the 350-500 Hertz range, and owls are all over the place
-In my experience, most good audio occurs between 10 p.m. and 4 a.m.

3 comments:

  1. That was a great post! Watch out, I might not need to send you my audio files anymore! I didn't know the spectrum option was in audacity.

    ReplyDelete
  2. It's not as good for detailed studies of audio, as say Sonic Visualiser, but it's good enough to do rapid audio file reviews. Once I find a good vocal in Audacity, I export it to a smaller clip, and then open it in SV for really close scrutiny. Unfortunately SV can't open huge files as well as Audacity, otherwise I'd just use it for all my audio file review.

    ReplyDelete
  3. Wish I had seen this sooner. Thank you for the info. Wondering if there is a way to load more than 20 minutes of audio to review at a time into Audacity. You mentioned the ability to load a file and then review in sections. I have not been able to open a large file (6-9hr) I am aware the vocals do not show up in Sound Organizer (in the 'line' profile graph), so I listen to many of the recordings in their entirety. There are other sounds of interest on the recordings besides the vocals, wondering if you have compiled a standard catalog with ranges. Any feedback greatly appreciated.

    ReplyDelete