Using Audacity to Get The Most Out of Your Recordings

Over the last few months I've had the opportunity to review a lot of "wild life" audio which normally involves digital recorders used to capture sounds over great distances. Often these recordings produce faint, noisy results that to the untrained ear possess nothing of interest. But after some looking around and a little practice, I've learned some techniques that work for amplifying weak sounds while filtering out noise to improve audibility. This has allowed me to go back and review hours of recordings that I thought were basically empty, only to discover an amazing amount of very interesting audio. This has proven so useful that I thought it worth while to pass along this information to anyone else who may want to give it a try. So here's my approach to getting the most out of my recordings using Audacity.

Here's what you'll need: the free audio editing program Audacity (download from: http://audacity.sourceforge.net/ , get the 1.3.x beta, it's stable and has the features needed for this work), a reasonably good computer (probably the one you're using now will be fine, I used a five year old laptop for most of my efforts), and a good pair of headphones (over the ear type are best, you don't want stray noise getting in the way).

After installing Audacity, here's the process I use to first amplify, then noise filter, and then secondary filter the recordings for best results.

Open the audio file you wish to edit in Audacity. If you have a long file, multiple hours, I recommend selecting/highlighting about 30 minutes of audio maximum to work with at a time (any more can take a long time to filter). After selecting the time frame you want to work on, click on "File" and then "Export Selection" to export your highlighted section to a new file, with a new name (this step is not necessary if you're working with files that are shorter than about 30 minutes).

Once you've opened the file you wish to edit, it will appear as a "spectrum" in its own little window. Go to the "Effect" menu and open it, then select "Amplify". The entire spectrum will be highlighted and a window will pop up. In that window it will provide a default Amplification(dB) and a New Peak Amplitude(dB). Normally it will automatically choose the amplification needed to achieve a new dB level of 0.0. This is often good enough for our purposes, and you can just click "OK" and be done. The whole file will be amplified. But you can also enter a larger or smaller amplification number and use the "Preview" button to see if you like the result better, then click OK and be done.

Next, we want to use the "Noise Removal" tool. First, click the little icon at the top of Audacity that shows a magnifying glass with a plus symbol in it. This will zoom you in one step for each time you click it. Zoom in until you can see the individual seconds of your recording, left to right across the top of the spectrum window. You want to find a spot in your recording that has the "noise" you want to remove, but has none of the faint background sounds that you want to keep. It takes practice, but you'll be able to spot these areas soon, they're normally the flattest part of the recording's spectrum. Once you've found that quiet, white noise spot in your recording, use your mouse to click and drag across a couple of seconds of that white noise. This will select that noise for use in the Noise Removal tool. Now, go to the Effect menu again and from there select the "Noise Removal" tool. A little window will pop up and at the top you'll see a button labeled "Get Noise Profile". Click it.

Next, on your keyboard, hold down the "ctrl" key and hit the "A" key (Ctrl-A for short). This will select the entire audio file again. Now go back to the Effect menu and select the Noise Removal tool a second time. In the section labeled "step 2" there are three numbers you can adjust. The Noise Reduction number is where you get the most bang for the buck. If it's a fairly quiet file you're working with, then a setting of 6 might be fine. If it's a very noisy file, you might try 12, 18, or even more. But be sure to use the "Preview" button in the lower left corner to see what your result will sound like. If you have the Noise Reduction number set too high, your result will have a "tinny" or metallic sound that is worse than the original noise. You want to select a number low enough to minimize that metallic sound, and high enough to minimize the noise in your recording. After setting the Noise Reduction number, you'll want to choose a number for the "Frequency smoothing" setting. I've always found 75 seems to be a good setting here. And finally, the "Attack/decay" setting, I normally set this to zero as my experiments have found no benefit to any other setting. Click OK and the noise will be filtered from your file.

Next step is to apply a "Low Pass Filter". Go back to the Effect menu, open it, and look way down at the bottom. Click on the Low Pass Filter. The settings I use most often in this window are 6 for the "Rolloff (dB per octave)" setting, the default (0.7) for the "Filter quality" setting, and 2000 for the "Cutoff Frequency (Hz)". Click OK and the frequencies above 2000 hertz will be increasingly reduced in volume by 6 decibels for every octave above 2000. The result is a recording with much of the high end hiss minimized. You might have noticed the "High Pass Filter" on the Effect menu as well. This works similar to the Low Pass Filter, but instead of letting everything lower than the cutoff frequency pass without modification, this will allow everything higher than your cutoff frequency to pass through. The High Pass Filter is useful for minimizing loud low frequency noise, like excessive wind noise on the microphone.

By this point, the Noise Filter and Low Pass Filters have probably knocked the overall volume level of your recording down a bit. So, as a last step, it might be advisable to re-perform the "Amplify" step described above. This should bring the recording level back up to 0.0 dB and make it suitable for basic listening on computer speakers.
Note that we haven't saved our work at any point in this process. Now might be a good time to to do so. But instead of saving it, you'll want to export it. Do this by selecting the "File" menu and then clicking on "Export". A box will pop up where you can enter a new file name (a good idea so that you don't over write the original file). Type in a new name and then hit OK.

As with anything, a little experimentation might improve your results. So feel free to play with those settings discussed above. But for a quick start to getting the most out of your audio files, these suggestions should work for most people.

Now you're done. You have an amplified sound file, with noise effectively reduced, and extraneous high end frequencies filtered out to improve audibility. Now give the file a play back through your head phones. Turn the volume up to a comfortable setting and you should be amazed at how much better the original audio sounds, and how many faint or previously inaudible sounds you can now hear clearly.

Progress Toward Launch

The basic assemblage of researcher contributed audio content has begun. Audio studies are being generated and organized into vocal classes. Much work remains however.

New Recording Collection

I've been doing a lot of review and analysis of the recordings I brought home last year. After some careful attention I've begun to find things that I had previously overlooked. A new collection of those recordings is in order, and found here:
https://sites.google.com/site/mongahela/myrecordings2009