Cleaning Up Frog Call Recordings with Audacity

Photography knowledge exchange.

Moderator: Scott Waters

Post Reply
User avatar
chrish
Posts: 3295
Joined: June 7th, 2010, 11:14 pm
Location: San Antonio, TX
Contact:

Cleaning Up Frog Call Recordings with Audacity

Post by chrish »

Thought I would throw together a quick how-to for cleaning up Frog Call recordings for entry into the NAHERP database.

The reason I think this is important is that unlike entering a photo of a frog in the database, a recording tells you more about the population, i.e. that they are breeding. Furthermore, you can record frog calls from private property which might be otherwise inaccessible for photos. However, getting recordings as vouchers into the database is a bit more difficult that just uploading a photo from your phone. So here's a simple and free way to get the recordings vouchered. It may seem convoluted here, but actually it only takes a few seconds once you are used to it, and many of the steps aren't necessary for each recording.

I should point out I am not an audio engineer and I have only the most basic experience doing this sort of stuff. Heck, I don't understand half the language of recording which may become evident in my misuse of it in this post! :oops:

The program I use for this is a free program called Audacity. There are other sound editors but Audacity is easy to use, powerful, available for all operating systems, and did I mention it was free?

You can download Audacity at http://www.audacityteam.org/. (updated link Feb 2016)
In order to export MP3s (which the database requires), you will probably also need to install the free LAME MP3 encoder.
You might also need the free FFmpeg Import/Export Library to import your audio files since by default Audacity only imports a few file types.

These add-ons work within Audacity so once you install them they work seamlessly.

There are great installation instructions for Windows/Mac/Linux on these links. If I can install this, anyone can!

So here's how I generally work on my calls for the database:

_____________________________________________________

1. The the first step is going to be to open your recording in Audacity. Once you open it, you should see a screen that looks something like this, showing the waveform of your recording (assuming you haven't changed the default startup window in Audacity).

Image

This represents the sound of your recording. The X axis is time and the Y axes are the relative amount of volume at that time. Therefore, each "blip" you see here along the line represents a loud frog call while the line represents the quieter times between calls. Note that this is a very clean recording and there wasn't whole lot of background noise which is why we see such contrast between the call and the quiet. If you have a recording of a loud chorus, you may not see any quiet areas.

Here's what the recording sounds like at the moment.



____________________________________

2. In order to better see what is going on in our recording, we want to look at the spectrogram. We do this by using the little pulldown triangle next to the (partial) file name in the waveform window. One of the choices that we will have is Spectrogram. Choose this.

Image

____________________________________

3. Now we see the spectrogram (sonograph) of our frog call recording. However, notice that there are two separate spectrograms or tracks. This is because our recording was made in Stereo. Your recording may not be stereo and so you might only see one track. If so, skip this step.)
There is no real value to saving the stereo version of this frog call because it doesn't add anything to the recording and we were recording a point source of sound (a single frog). So we can combine both tracks into a single monoaural track to save file size.

To do this, go to the Tracks menu and choose Stereo Track to Mono.

Image

____________________________________

4. Now we have one track, but it is only occupying a small part of the Audacity window. We want to be able to look at our spectrogram, so we want to expand it to fill the screen. You can do this by going to the View menu or simply by clicking Ctrl+F to fill the width and Ctrl+Shift+F to fill the height of the window.

Image

____________________________________

5. Our next step is to read our spectrogram.

Image

We can see that the frog made 7 discrete calls during the 11 second recording.
We can see that the main call (indicated by the red arrow) starts off somewhat quietly and slides up in pitch to the loud finish. This is why it is the shape of a quotation mark. The thin area is quiet, then it goes up (increases pitch) and becomes darker (increases volume) until the more complex end note. The blue arrows represent higher pitched harmonics of the main call.

But if you look at the scale to the left, you will see that although the main call is somewhere between 2000 - 5000 Hz in pitch, the harmonics are much quieter (lighter colored) and much higher in pitch (one is at around 11000 - 12000 Hz and the other at 14000+ Hz). Normal human hearing ranges from 20 - 20,000 Hz. But sounds above 12000 Hz are very high pitched and hard to discern unless they are loud. So in reality, this higher pitch harmonics are not audible over the loud, lower pitched call (to most people).

In case you don't believe that part of the call doesn't matter, here's what it sounds like. I have removed the part of the call below 10Khz and then amplified the remaining call harmonics. See if you can hear them or if you think they matter. (WARNING - this recording is very high pitched and somewhat uncomfortable to listen to...sort of like Mariah Carey :lol:).



So we don't really need all that frequency "headroom" above about 8000 Hz because there isn't any signal (call) there and what is there is just background noise. So we are going to reduce the Project Rate to cut out those extraneous high pitched and ultrasonic frequencies.

_____________________________________

6. Reducing the Sample Rate

We can use the pulldown menu in the lower left hand corner to reduce our project sample rate from 48,000 to 32,000 Hz. I actually ended up at 16,000 Hz.

Image

How do you decide? Theory says you can record a frequency of one-half of the sample rate. Therefore, if you are using a 32000 Hz recording rate, you can only record frequencies up to 16Khz. Why do this? It can reduce the filesize of your final recording.

This can be very helpful when recording frogs in areas with lots of insect noise. Many insect calls are very high pitched and you can filter those high pitched calls out of a recording by allowing the frogs to be better heard. If I have a recording with a katydid calling constantly at 11,000 Hz and my frog is calling quietly at a lower pitch of 3500 Hz, I can reduce my sample rate to 16000 Hz and the katydid call is gone from the recording.

The great part about audacity, is that changing the project rate doesn't actually change your original file, it only changes the output you export from the file (which we will do later). So you can play with this value to see how much headroom you really need. Try using double the apparent highest frequency of the frog's call. Fortunately, when you change the Project Sample Rate with this box on the bottom left, Audacity now plays the recording at that new sample rate so you can hear what difference it will make by pressing the Play button on the top of the page. (NOTE: The Play button changed to that funky green symbol when I pressed Ctrl-PrtScn to make the screen capture. Yours should look like a normal triangular Play button).

Listen to what it sounds like. Do you hear a difference? If not, try something lower. So I might try 32000, then 22050, then 16000 then 8000 to see how much I could "trim off" from the top frequencies. When I did this for this recording I found that I couldn't really hear much difference between 48,000 and 16,000 but when I went down to 11,025 Hz, I heard a change in the "character" of the call. This tells me that there were important aspects of the call between 5512 Hz and 8000 Hz, but nothing important above 8000 Hz. So I chose 16000 Hz as my Project Rate.

Here's that original call at 48Khz:



and the same recording resampled at 16Khz



______________________________________________________

7. Removing low frequency noise

You may have noticed in the recording that there is a lot of background noise. At the end of the recording, you can actually hear a car approaching on the nearby road. Fortunately, these man made noises are generally fairly low frequency and below the minimum frequency of this particular frog call. Therefore we can filter out those low frequency noises without losing frog calls. This isn't always the case however. Some Ranids have deep resonant calls that overlap these human made noises so this won't work in all cases. So we need to look and see how low our calls are and how much we can remove below that level.

You might be able to do that by looking at the spectrogram as is, but in some cases it is helpful to zoom into to the lower part of the frequency axis. To do that, click the mouse on the lower part of the Y axis (where the red arrow is pointing). In windows, left clicking zooms in and right clicking zooms out.

Image

So now I am only looking at the part of the spectrogram between 0 and 6Khz. What I can see is that my frog's call is above 2Khz. I have added an orange line to show that level and how I can tell that. Therefore I can safely remove any noise below 2Khz without losing any of my frog call. To do this, I am going to use the High Pass Filter Tool from the Effects Menu.

______________________________________________________

8. High Pass Filtering

In high pass filtering, you basically define a minimum frequency for your recording and the filter removes any data (noise) below that frequency level. We decided that we can safely go somewhere below 2Khz, but for safety, we are going to try about 1800Hz (1.8 Khz) instead to avoid losing any of the lowest parts of the frog's call.

Image

_______________________________________________________

9. The High Pass Filter Menu

In the high pass filter popup, you will see a couple of choices to be made. The first is the Rolloff. This is a measure of how much filtering you want to have done; i.e. how much sound you want to remove. 6 db removes just a little amount, 48 db removes the most.

Image

Your knee jerk reaction might be to say "well, I want to remove ALL the extraneous noise". But if you silence the recording below that level, it seems to call more attention to the noise above that level. I'm not sure if that is mathematically or physically true, but the recordings can sometimes sound "hissier" if you overdo this. I usually try about 24 or 36 db, see how it sounds and then undo/redo it 'til I get it where I like. NB - it is possible to remove all of the lower frequency sound and then put in some "cleaner" low frequency sound ("brown noise") to make the recording sound natural again. Here is a website describing how to do that in a different program, but Audacity will also generate brown noise if you want to add it back - http://naturesound.org/?page_id=465.

EDIT - I have subsequently found it is better to try and do this in small steps. Try 6 db at 500Hz and listen to the result. If you want more, you can always hit CTRL-R to run the same High Pass Filter again. If it is too much, CTRL-Z will undo it.

The other option to adjust here is the cutoff frequency. This is the frequency threshhold for your filtering. If you set this at 2000 Hz, it will remove sounds below that level. I think 1800 Hz is a good safe level for our recording that will remove noise but leave all of the call frequencies. For some species, like the larger Ranid frogs, their calls are at low frequencies and require a lower cutoff value.

Image

Here's what our spectrogam looks like after the high pass filtering:

Image

Here's the recording before High Pass Filtering:



And here it is after High Pass Filtering:



Notice the car at the end is effectively gone.

________________________________________________________

10. Amplifying the sound

Now that we have removed some of the excess noise, we may want to amplify the sound. This step may also be performed after the Noise Removal step in any recordings with a lot of background noise. Go to the Effect menu and choose Amplify. There are several ways to do this, but we'll use Amplify for this example. When you choose Amplify, the Amplify dialog box pops up.

Image

There are several things to set:

The Amplication window tells you the amount of amplification that will be applied. Here, the default setting will increase the volume (technically the sound pressure level) by 8.2 decibels.
The New Peak Amplitude window tells you how loud the loudest part of your recording will be at that amplification setting. 0.0 is the maximum Peak Amplitude you can have without distortion (or clipping) occurring. Negative numbers are below that level, positive numbers are above that level. The slider between them adjust the Amplification and Peak Amplitude together since they are tied.

By default, Audacity will not allow you to increase the Amplification to any level that will cause distortion and loss of information (clipping). If you wanted to increase beyond that level, you would have to choose the Allow Clipping checkbox.

The default value for Amplification when you open the tool is at the maximum value you can use without clipping. I find I prefer the sound if I use a level below that maximum. Fortunately, there is a preview button which will play the effect of your settings before you committ to them by clicking OK.

It is sometimes more useful to be looking at the waveform of your call when doing the amplification. You can get back to that view using the little pulldown triangle next to the abbreviated file name in the file window corner.

Here's the recording before amplification. The 1 and -1 on the Y axes refer to the maximum peak without distortion. You can see that we are less than 40% of the way to the maximum.

Image

The default setting for the Amplify dialog box is to increase the volume until the loudest peak just reaches the clipping threshhold. So here's the waveform from using the default values:

Image

If you were to raise the Amplification too far, you would get clipping which looks like this on the waveform:

Image

And sounds like this:



EDIT - I now usually keep the amplification to set the peak value at something like -3.0 db. You can do that by setting it in the bottom window, or just subtracting 3 from the suggesting amplification value (i.e. if they suggest 21.4, enter 18.4).

___________________________________________________________

11. Cutting out non-standard peaks

(Note, this step is also useful in editing out most of the intervening noise, recordist noise or things you don't need in your voucher. Remember that a voucher recording only needs to be long enough to identify the species in question. Three minute recordings simply take up disk space on the NAHERP server without adding anything that couldn't be accomplished with the appropriately edited 10 seconds.)

Looking at the waveform, you can see that the first call in the series is much louder than the others. Because of this, it effects the maximum amplification I can use.

Image

Therefore, if I removed that one call, I could amplify the others and keep then all at the same level. To cut/crop a section of the call, I simply click and drag in the recording window and select (it gets darker) a region of sound. Then I can either hit Delete on the keyboard or click the Scissors button above the recording window.

Image

I actually trimmed off the first two calls and the last little noise at the right end of the waveform. Here's what I ended up with after trimming the loudest calls and them amplifying.

Image

For this call, I chose an Amplification of 5.2 which gave me a new Peak Amplitude of -3.0.

Before:



After




Now, theoretically, I could be done. I could export an MP3 from this recording and upload my new record to the NAHERP database. But you might not like all the extra hissing and background noise in the recording. Let's see what we can do with that.

________________________________________________________________

12. Noise Removal

For noise that was above or below the frequencies of our calls, we simply used High Pass Filtering and a lower Sampling Frequency to remove/minimize it. However, for noise that is within the range of our calls, we can't do that. In this recording, the wind was blowing through the vegetation all around me and the microphone is picking up that extraneous wind noise as well as other noise. To get rid of this, we can use Audacity's Noise Removal tool.

The way the noise removal tool is going to work is we are going to show Audacity what the background noise sounds like and it will then try to remove that noise from "around" our signal (calls). This is most easily done viewing in the Spectrogram mode.

We will first choose a section of the recording that contains only noise and no calls. We choose this like before by selecting a few fractions of a second using the mouse to click and drag define/select the section. It is a good idea to press Play after selecting this section to make sure there is nothing in there you don't want removed, like a quiet background call.

Image

Once that has been selected, then we open the Noise Removal dialog box (under the Effects menu). When it opens, and we have our noise-only section selected, we click Get Noise Profile.

Image
Image

When you click this the dialog box "captures" the noise profile then closes.

Now you want to go back to your recording and select the whole thing by double clicking in the recording somewhere.

Image

Now reopen Noise Removal and the dialog box reappears. This time you want to adjust the settings of the lower sliders before hitting OK. The defaults are probably good for starters. You have to be cautious here because some of the "noise" that might be removed could be part of the calls.

Here is the resulting spectrogram after noise removal:

Image

Notice the noise removal process appears to have added some low frequency noise below each call. Repeating the High Pass Filter with the same settings as before will clean that up.

Here is the result of the noise removal. Some data from the call itself has been lost, but the call is still clear and diagnosable as a voucher.



________________________________________________________________

13. Exporting the final mp3

Now that you have the product you want, it is time to export your mp3.
For that you use the File:Export dialog.

Image

If you are exporting the whole recording, choose Export.
If you are only exporting a cleaned up selection of the recording or the call of one species from a chorus recording, select that section and choose Export Selection.

Choose MP3 as the file type from the pulldown screen since this is the file type that the NAHERP database will accept as a voucher. If you don't see the MP3 option, you forgot to install the LAME MP3 encoder mentioned at the top.

Image

After naming the file a dialog will open allowing you to enter tags for the MP3.

Image

Note that you can save a template. This would allow you to save certain settings that you always use for frog mp3s and you wouldn't have to type it again each time. I have two saved templates, one for Anuran calls and another for Bird Calls. By saving them as templates I can ensure consistency in my MP3 tags.

After that, you can simply upload the file as you voucher into NAHERP.

Of course, the one I am working here with is a recording of Eleutherodactylus johnstonei from Curaçao, and that species doesn't occur in the range currently covered by the NAHERP database. That will be fixed whenever the WORLDHERP database gets set up! :lol:
EDIT - an now it is - HERPMAPPER.org and this frog can be seen and heard here - http://www.herpmapper.org/record/6331

Thanks for being an NAHERP contributor! :beer:

Chris
User avatar
tomharten
Posts: 130
Joined: June 21st, 2011, 11:37 am
Location: Southern Maryland

Re: Cleaning Up Frog Call Recordings with Audacity

Post by tomharten »

Thanks for posting this! We're in the second year of a middle school science program in which we are having students collect frog and toad sounds from their backyards and school yards. We are also using Wildlife Acoustics SM2 automated recording systems to capture sounds throughout the night (usually in one minute samples per hour). I'd like to play around with Audacity with the students, in part to clean up some of the recordings, but mainly because I think the kids would find this very interesting. Our data is being contributed to the Maryland Amphibian and Reptile Atlas project and I would be interested in having it go to the NAHERP database as well.

Thanks again!

Tom
User avatar
Tom Lott
Posts: 43
Joined: July 12th, 2010, 3:29 pm
Location: south-central Texas
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by Tom Lott »

Chris,

That's an EXCELLENT tutorial. Thanks for taking the time to put it together and posting it; it will doubtlessly save me a few hair-pulling sessions!
User avatar
crocdoc
Posts: 473
Joined: June 7th, 2010, 11:43 pm
Location: Australia

Re: Cleaning Up Frog Call Recordings with Audacity

Post by crocdoc »

Excellent tutorial and very useful!
User avatar
Kevin Messenger
Posts: 536
Joined: June 7th, 2010, 3:00 pm
Location: Nanjing, China
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by Kevin Messenger »

thanks for this!! just what I was looking for!
User avatar
chrish
Posts: 3295
Joined: June 7th, 2010, 11:14 pm
Location: San Antonio, TX
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by chrish »

Kevin Messenger wrote:thanks for this!! just what I was looking for!
I am actually thinking about updating this since I have a few more thousand calls worth of experience, but the main stuff is the same.

Some things I have learned over the last year or more are -
- I often prefer to use the Equalizer rather than High Pass Filtering
- that the Parametric Equalizer is a great tool
- and I usually don't amplify beyond a peak amplitude of -6db, depending on the nature of the call
User avatar
Kevin Messenger
Posts: 536
Joined: June 7th, 2010, 3:00 pm
Location: Nanjing, China
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by Kevin Messenger »

I'd love more opinions/ skill on drowning out stream noise. I was able to reduce the noise, but still not 100% happy with it. Where did you upload your audio so that you could embed it? Flickr?
User avatar
chrish
Posts: 3295
Joined: June 7th, 2010, 11:14 pm
Location: San Antonio, TX
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by chrish »

Kevin Messenger wrote:I'd love more opinions/ skill on drowning out stream noise. I was able to reduce the noise, but still not 100% happy with it. Where did you upload your audio so that you could embed it? Flickr?
Stream noise is tough. It is even tougher than road noise because it occurs across a wider range of frequencies. You can't just reduce the gain of a low frequencies to get rid of it.
The noise reduction tool may help with this, although I'm never 100% happy with the results as you lose some of your signal.

If you signal (call) has a fairly narrow frequency range you could use the equalizer or Parametric Equalizer to bring that up over the other wavelengths, but if the signal has a broad frequency range it might be tougher.

I have embedded audio files on these forums from
- HERP
- Herpmapper
- my own website
- Dropbox

Dropbox is probably the easiest free way for files you don't want to put on HERP/Herpmapper.
User avatar
chrish
Posts: 3295
Joined: June 7th, 2010, 11:14 pm
Location: San Antonio, TX
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by chrish »

Here's an example where I used the equalizer to bring out a frog's call from the noise of the stream. Of course, it wasn't a particularly overwhelming level of stream noise because I had used a shotgun mic so it was pretty clear in the original "noisy" recording.

This is Litoria rheocola from along a rocky stream in Queensland.





And here's the result where I used the equalizer to reduce the gain (volume) of everthing except the 2300-3300Hz range which is the range of the call.





BTW - those are housed on dropbox.

Chris
User avatar
Kevin Messenger
Posts: 536
Joined: June 7th, 2010, 3:00 pm
Location: Nanjing, China
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by Kevin Messenger »

nice, thanks, I'll give it a shot. My stream is way louder, but I think it can still be improved upon. I'll try to put the original on dropbox and see if I can embed it

to embed on here also has to be mp3 I guess?


User avatar
chrish
Posts: 3295
Joined: June 7th, 2010, 11:14 pm
Location: San Antonio, TX
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by chrish »

I downloaded your call and tried to clean it up a bit more. It is a tough recording because the call has a fairly wide frequency range (~2500-5000hz) with a carrier frequency at 3513 (according to Audacity). But the background noise is also strong in these frequencies as well. However, a good part of the noise is below this and can be reduced.

Here's a quick 2 minutes cleaning.
All I did was reduce the sample rate to 11025Hz then ran it through an equalization to reduce the frequencies outside the range of the call (2500-5000hz).
I ran the equalizer tool twice at these settings -

Image

Still not perfect, but the call stands out quite a bit more I think?



Noise reduction might be a good tool to try to this as well.

_________________________________________________________

OK, so I tried something else. This time I treated different parts of the same recording. Since there are 8 calls on the recording, I treated each pair of two differently.

1. First I reduced the volume (Amplify tool) to -6.0dB. I did this so that I could standardize the volume of each section after treatment.

2. I left the first two calls as you had them.

3. The next two calls (calls 3 and 4) were treated as follows -
a. I selected some of the water noise and used to filter out using the Noise Removal Tool (removal set to 18dB).
b. After noise removal, I ran the same Equalization profile from above
c. I then reamplified what was left back up to -6.0dB to make them comparable.

4. The next two calls (calls 5 and 6) were just treated with the Equalizer as the previous recording, but with no noise removal first (and then reamplified to -6.0dB)

5. The last two calls were unchanged from your recording.



I think I like the noise reduction + equalization calls the best. I hear more frog (?) and less background noise.
User avatar
Kevin Messenger
Posts: 536
Joined: June 7th, 2010, 3:00 pm
Location: Nanjing, China
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by Kevin Messenger »

very interesting and what a unique way to compare techniques and styles! I've been saving multiple copies and exporting: v2, noise reduced, v3 noise reduced and resampled at 11000 hz, v4, X, and Y, and Z, and a myriad of combinations! and then trying to introduce the equalizer as well. It makes it difficult to compare across the board. But I like how you did it a lot.

So when I was looking into audio programs, I found Raven Lite via cornell university. It gave me a great distribution of the sonogram as well as the frequency:

ImageXeno sp nov, cleaned, full length by Kevin Messenger, on Flickr

(also in the above version, I essentially used Raven Lite to remove the sound in between the calls - that's why it looks clean, but it's not very "natural" if you were to listen to the audio attached to the image.

But I'm curious, earlier you mentioned that the khz were 2.5 - 5, whereas I interpreted the range as 3-4.8 (I need the range for scientific reasons). So were you just doing a quick eye-balling, or are you seeing something I'm not?
User avatar
chris_mcmartin
Posts: 2447
Joined: June 9th, 2010, 12:13 am
Location: Greater Houston TX Area
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by chris_mcmartin »

This is an awesome topic; can't believe I missed it when it debuted (but then again, it's in the Image Lab... :lol: ). I'll have to give these cleanup techniques a try the next time I record calls.

I use Adobe Audition for most of my audio-editing needs, but regarding amphibian files mostly it's been to cut extraneous bits from the timeline and to amplify the calls. I didn't even think of using a sonogram-type display, but because of this thread I explored how to do that in Audition. It's "View-->Show Spectral Display."

Chrish, in case you haven't seen it, there's a whole birding-related blog dedicated to recording, editing, and interpreting calls.

http://earbirding.com/blog/

A lot of the techniques should easily translate to amphibian calls as well. :thumb:
User avatar
chrish
Posts: 3295
Joined: June 7th, 2010, 11:14 pm
Location: San Antonio, TX
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by chrish »

very interesting and what a unique way to compare techniques and styles! I've been saving multiple copies and exporting: v2, noise reduced, v3 noise reduced and resampled at 11000 hz, v4, X, and Y, and Z, and a myriad of combinations! and then trying to introduce the equalizer as well. It makes it difficult to compare across the board.
Another useful way is to get a short clean section of calls from your recording and just copy paste that into a new file 4 or 5 times in a row to test different treatments. Allows for easy tweaking. It helps to leave a 1/2 second of silence or so between the copies to let your ear "adjust" to the changes.

You can also cut and paste to "improve" a call that is messed up by background noise. Yes, it is cheating, but I see it like cloning a section of a photo to cover a beer cap next to your coiled snake. That's what I did here to compare the two Acris calls. The first 5 crepitans clicks were 5 real-time consecutive clicks. But I took a single gryllus click and cut and pasted it so that its rhythm would be identical to the crepitans thus accentuating the differences in the length of each click.




Kevin Messenger wrote:But I'm curious, earlier you mentioned that the khz were 2.5 - 5, whereas I interpreted the range as 3-4.8 (I need the range for scientific reasons). So were you just doing a quick eye-balling, or are you seeing something I'm not?
No, that was just a quick eyeball boundary. You more boundaries are more precise, although it looks like there is some signal below between 2.9 and 3Khz.

The carrier frequency I got from Audacity's Plot Spectrum graph (under the analyze menu). It shows the carrier frequency (peak frequency) of around 3.5Khz. If you are writing this up, that is a useful datum to include. When going over my Costa Rican frog recordings, I found Jay Savage's inclusion of this datum invaluable in distinguishing similar sounding calls.

I actually did that in this recording when I was trying to compare the two Acris species calls. The crepitans (first 5) is a real recording but the gryllus is a single click call repeated five times.

Kevin Messenger wrote:So when I was looking into audio programs, I found Raven Lite via cornell university. It gave me a great distribution of the sonogram as well as the frequency:
chris_mcmartin wrote:I use Adobe Audition for most of my audio-editing needs
I actually have Raven Lite and Audition as well as Audacity.

I agree that Raven Lite's side by side spectrogram/waveform view is nice. In Audacity you have to see one or the other unless you open two copies of the same file in the window.
But I find Audacity easier to use than Raven Lite for the same sort of processes. I started with Raven Lite and found I wasn't getting very far. Then I downloaded Audition and found it more user friendly.

As for Audition, I downloaded Audition for "free" a few years back when Adobe screwed up and offered their older software for "free" for current owners.
I have tried to learn it on multiple occasions but always find myself saying "hell, I could do this in 30 seconds in Audacity" because I know Audacity.

I think comparing Audition to Audacity is like comparing Photoshop to Lightroom. Yes, Photoshop does give you infinitely more control, but Lightroom gives me enough and is simpler to use. Most of the tools in Photoshop and Audition I would never use. Then again, I haven't deleted from my computer and I keep telling myself to learn to use it properly.
chris_mcmartin wrote:http://earbirding.com/blog/
Absolutely, earbirding is on my "frequently checked blogs" list.

Some other favorite recording sites -
http://www.avisoft.com/tutorial_field_recording.htm
http://txbirdsong.me/ - it is about birds, but offers some good tips sometimes
I could go on, I have a whole folder on my bookmarks with dozens of recording techniques and gear links

Lang Elliot's blog isn't really a "how to", but it is a great blog for people intersted in anurans (and other sounds)-
http://miracleofnature.org/blog/category/amphibians

Chris
User avatar
Kevin Messenger
Posts: 536
Joined: June 7th, 2010, 3:00 pm
Location: Nanjing, China
Contact:

Re: Cleaning Up Frog Call Recordings with Audacity

Post by Kevin Messenger »

I will be writing it up - as we think (pretty, pretty sure) it's a new species of frog.

Again, thanks for all the assistance and the tutorial, incredibly helpful and I'm glad someone in "the group" has this sort of skill (however partial of a grasp that person may feel they have on it) such that they can teach it.

I was so green that I didn't even know the first steps into what or how to go about cleaning up the noise. The best I could do is the noise removal and even that I later discovered I was doing wrong.

Interesting that you find lightroom easier than photoshop! I find it the reverse. Lightroom is too intimidating, photogshop seems way more direct and straightforward.

I also started with raven light, but then downloaded audacity. and aside from the measuring and looking at the spectrograms and waveforms, I see no benefit raven lite has over audacity
Post Reply