• BirdForum is the net's largest birding community dedicated to wild birds and birding, and is absolutely FREE!

    Register for an account to take part in lively discussions in the forum, post your pictures in the gallery and more.

best sonogram/spectorgram software out there? (1 Viewer)

ermine

Well-known member
Cornell Raven

http://www.birds.cornell.edu/brp/raven/Raven.html

seems to be the de facto standard. But not free, which is curious. I was under the impression the US had an enlightened approach that mandated that information collected or developed on Uncle Sam's dollar using public funds had to be put in the public domain. Obviously not in this case .

if what you meant was best free software ;) try Christoph Lauer's Visible speech

http://www.christoph-lauer.de/

which is a Java app. It presents a heavy load for your PC

I rather like the unassuming spectrogram - the freeware version is

http://www.hitsquad.com/smm/programs/SPECTROGRAM/

and payware version is

http://www.visualizationsoftware.com/gram.html

this is much easier on your PC, though nowhere near as comprehensive as Christoph Lauer's one

http://www.nauta-rcs.it/SEA/seawave_index.html may be worth a look

Actually it looks like I may have been unfair to Cornell, Raven lite appears to be free now :)
 

griffin

Well-known member
Steelflight said:
anyone know what the best sonogram/spectoram imaging software out there is for bird calls?

Used to use Gram23 which was okay - easy to standardize files, but poor contrast control of images - they can look too faint.

Raven Lite ( harder to set/standardize axes). Good if it is now free though at $25 was hardly a rip off.

Now use Raven Pro 1.2 and am happy with it, but $400.

There is another called Praat, but like Spectogram the learning curve looks steeper.

Linz
 

ermine

Well-known member
griffin said:
There is another called Praat, but like Spectogram the learning curve looks steeper.

Linz

Did you ever work out what the heck Praat was telling you? I couldn't make head or tail of what I got out of it. The results bore no resemblance to anything I had seen elsewhere. They did at least vary with the input signal...
 

g8ina

He's pining for the Fjords !
Im using (have stopped using actually, cuz I dont do bird recording any more) Cool Edit Pro. It has spectrograph capabilities, but it sure not a freeware job :)

My samples are in my 32" parabolic dish for sale link.
 

griffin

Well-known member
ermine said:
Did you ever work out what the heck Praat was telling you? I couldn't make head or tail of what I got out of it. The results bore no resemblance to anything I had seen elsewhere. They did at least vary with the input signal...

Afraid not ! Played with it for a bit but felt I would need to invest a helluva amount of time to get somewhere with it !
 

lachlustre

Should be recording bird song
It's a shame you guys didn't like Praat - I love it myself. It is by far the most powerful program - in terms of analysis - out of the ones metioned here. That means that it can be a bit daunting to being with, but I didn't find it too weird myself.
Praat also has a really large community behind it in the speech world, and the programmers have an excellent reputation. One good thing that results from this is that the code is pretty solid at this point. I've found it a lot *less* buggy than the other programs listed here. We had some unpleasant problems, for example, trying to resample wav files in CoolEdit a few years ago.
As some of you know, I'm working on my own bird song software. So a question for all of you: what features would you all want in the perfect spectrogram program?
 

lachlustre

Should be recording bird song
By the way: I think Raven, like "Visible speech" is written in Java. If the latter presents a heavier load for your computer, then it's not because it's written in Java.

In the bird song research world, the following are the most-used programs. First the more traditional programs:
1) Raven - as described above
2) Canary - the mac precursor to Raven from Cornell (still liked by many)
3) Avisoft - it's 800 Euro because it keeps a one-man programming company in Berlin afloat. It's kind of cool if you can afford it, because he'll often incorporate features on request once you've paid for it.
4) SIGNAL - (Engineering Design) is the grandaddy: used to require its own hardware board when I used it. Modern version doesn't, but was still really expensive the last time I looked. VERY flexible, but not for the faint-hearted.

And here are the more recent programs that offer something a bit different. I would think that the number of publications based on the list below in the last few years is substantially larger than the number of publications based on the list above. Reflecting changes in computer hardware, and university reaserch interests, these are all university projects coded by individual or small teams of scientists. And they're all free! I honestly think that these might be useful tools to you guys. The documentation is sometimes (but not always!) a little more inaccessible than the commercial software, but people seem much more aware of this than they were a few years ago. My final reason why you should check these out: you've all paid for them - these are all research projects developed with government funding. Why not see what you got for your money?
5) Praat - see discussion above. This program, compared to those below, offers the most flexibility in manipulating sounds, and measuring esoteric aspects of it.
6) Sound Analysis Pro (http://ofer.sci.ccny.cuny.edu/html/sound_analysis.html) - this is a fantastic project by Ofer Tchernichovski at CUNY (New York). It's mostly aimed at zebra finch song research in the lab, and many of the tools in it are not much use unless you are doing that. The advantage to us field recordists is that this probably produces the most beautiful spectrograms of anything out there. Why? Two reasons: first is that it uses more advanced algorithms than your basic FFT (multi-tapering), which I think gets you better resolution; second is that you can plot spectral derivatives (change in spectral power over time) which turn out to be a great way of visualising sounds - check out the examples on the website.
7) Syrinx (http://syrinxpc.com/). I'm kind of surprised this hasn't been mentioned, because it really has been developed with field ornithologists in mind. I'm not really much of a user, because I use macs right now, and this is PC only (despite my nagging the author!). But what I have seen makes me think this is a gem. It's quick and easy to use, and also offers some neat recording options if you connect your computer straight to a mic. Essentially it provides an automatic recorder of bird sounds, which might appeal to those of you that might be recording birds at your feeder.

Hopefully, I can add my own program to this list within the next few months!
 

griffin

Well-known member
Hi Rob,

Maybe you need to give all us leckies a Praat masterclass/tutorial for bird call analysis ? ;)


Linz
 

ermine

Well-known member
That's a great rundown of some new programs to me - keep us out of mischief for a while!

The problem I had with Praat probably started with me. I have no background in biological sound research though enough in engineering. I fired Praat up and all that I could get our of it was what looked to me like a bewildering array of horizontal pink wavy lines across the screen. Reading the rest fo the website I saw this was heavily aimed at human speech, so I assumed it was telling me stuff that didn't mean much in terms of birdsong. I certainly found nothing that would give me what I recognised as a standard spectrogram. But that says more about my own lack of background - I may revisit it since you say it has value in this field.

As far as what I look for in a a spectrogram program is probably coloured by this lack of background. Take the sparrows in my backyard - everybody says that sparrows all sound the same and have one chirp. And yet it is clearly audible that the sound of a sparrow arriving before the flock and issuing a chirp clearly has a different tone colour from the same sparrow in the same location once the others have arrived. I would expect programs to be able to show differences like that. In fact I'd like them to help me ID the sparrows as individuals without ringing, I am trying to see if I can learn to hear this but so far with little success.

At the moment i find such programs sometimes show differences in some sounds that I can't hear (the differences, not the sounds), and sometimes don't show differences in sounds that I can hear. I'm reasonably familiar with the concepts of what they are doing and have used RF spectrum analysers so the concept of frequency resolution etc aren't totally new to me. But clearly I have much to learn in using this kind of app.

Spectrogram software also doesn't have anywhere near the discrimination against background noise that we have, and display resolution in the 500-1kHz range seems to be poor. Again, that could be operator error...

By the way: I think Raven, like "Visible speech" is written in Java. If the latter presents a heavier load for your computer, then it's not because it's written in Java.

Raven also revs up the CPU fan and whacks up the CPU load for me, both at home and at work. I really HATE java programs - they take an age to lumber into action and hog resources something rotten :C
 

lachlustre

Should be recording bird song
Ermine:

Your post quite accurately gets at some of the limitations of spectrograms. The more perceptual research is done, the more obvious it becomes that vertebrate auditory perception is quite unlike spectrographic representation. Ironically, one of the features of Praat is to try to derive some more perceptually relevant features (like pitch and formants) that give us a better picture of what's going on.... of course there are some differences between human and avian perception, and the algorithms used to derive these features work better with human speech than with bird song.

Neverthless, this means that your ideal spectrogram program, which reflects your perception of sounds better, would likely resemble Praat to some extent!

Another important limitation of FFT-based spectrograms is that there is a fixed trade-off between frequency and time resolution: you get the same frequency resolution between 500 and 1000Hz as you do between 5500 and 6000Hz. Well, it turns out that vertebrates perceive frequency in a logarithmic way. So that means that FFT's tend to show too much frequency resolution and higher frequencies and too little at low frequencies. Maybe this is what you were getting at when you mentioned problems at lower frequencies.
There has been sporadic interest in wavelet-based spectrograms which use a completely different set of algorithms, and can get around this problem... but it hasn't really amounted to much yet, I'm afraid.

I'll try to contruct a post that gets you started with Praat sometime. It's really not that hard to get going, but some of the default settings, which are indeed more tuned to human speech, should be changed.
 

ermine

Well-known member
lachlustre said:
Neverthless, this means that your ideal spectrogram program, which reflects your perception of sounds better, would likely resemble Praat to some extent!


I'll give it another go. Although I would expect a program to show me differences in tone colour that I do hear, naturally the object of the exercise is also to pick up the elements I don't hear :)

For instance, anybody who has done any editing will know that what we hear at any point is very badly influenced by what goes before, and for some reason I don't understand, what is about to come. You can't pick out a flaw and listen to it in isolation - any segment of say 50ms or less sounds like a click. And yet the same 50ms put in a longer piece of say 2secs duration is easily audible as being present or not, and takes on a character of its own.

Slowing down a wren's trill seems to show much more detail that I hear in real time, so our auditory temporal resolution must also really stink. I'm not so sure that detail is visible in spectrograms when I make them.
However, I can at least make my spectrograms look similar to Don Kroodsma's ones from his CDs in "the singing life of birds" and get something similar to the "Sound approach" guys so my settings can't be too atypical.

It's a pity these apps have no understanding of stereo. Some of what makes bird recording special is the interaction between individuals where they set each other off. and with a decent open mike recording one can easily place the individuals in stereo.
 

Users who are viewing this thread

Top