Dear Laura,

I am afraid your way ahead is a long way considering your approach. I must say that its a brilliant idea and I am myself interested in it. Though, you should have a good background in artificial intelligence or electrical engineering to do sound analysis.

I guess you have neither of these back grounds. What you plan to do requires elimination of the surrounding noise and extraction of effective features from the sound. Your work is close to disciplines such as Music Information Retrieval (MIR) as well.

I have carried-out several works with my student on music and similarity of the sound of the car engine to detect its malfunction. The latest paper is going to appear in the following link:

http://journals.tubitak.gov.tr/elektrik/index.htm;jsessionid=205D7D29F43F9311E4C805C683617DE2

at the following place:

http://online.journals.tubitak.gov.tr/openInPressDocument.htm?fileID=824004&no=154796&fileType=Report Document

You cannot carry-out your work using the off the shelf software.

Best

Peyman

Hello Laura,

Regarding your first question, that of dealing with stereo recording, your treatment and analysis would depend on several factors.  I am assuming that you used two microphones:  Were these the same model, calibrated to the same sensitivity, and oriented in the same direction?  If not, you would be better off keeping the recordings separate and treating them as independent samples.  Even if the answers to the above question are yes, you need to be careful about combining sound recording parameters.  Assuming that the microphones, sensitivity, and orientations are identical, two recordings - taken at the same time - could be combined to provide a more accurate measure.  Temporal parameters could be averaged, but intensity needs to be treated carefully:  A voltage reading (taken directly from the recorder) is a linear measure, and these can therefore be averaged.  But if the readings are in decibels, you would need to convert these measures to pascals, average the pascals, and then reconvert back to decibels (which are logarithmic measures).  As for your specific questions on sound ruler, I'm afraid that I cannot help here, as I don't use the software.

Hope this helps,  Michael Greenfield    CNRS     Tours, France 

Hello Dr. Greenfield,

First of all, thank you for your answer.

I recorded with a Zoom H4n Handy recorder (image attached), and as you can see, it had two internal microphones. Thus, they were the same model and they were also calibrated to the same sensitivity, but weren't oriented in the same direction. However, when I introduce the recordings into the software, oscillograms and spectrograms of both channels seem identical... Today I've read about Audacity software, and it allows to convert stereo recordings to mono ones. It combines both of them and reduces its volume in -6 dB (halves the volume) to obtain an average of both channels. Maybe it could be the solution to this problem, couldn't it?http://alphamanual.audacityteam.org/man/FAQ:Editing

Thank you again, and thanks to Dr. Kabiri too.

Best regards,  

Laura

Hello Laura,

1) If you want to describe different types of calls, I think you can safely use just a single channel. Differences between two channels are not important regarding your question. Choose the left or right channel (always the same). If you want to be sure. You can try to measure each of the channels separately for a random subsample of calls and correlate those values to provide evidence that they are indeed almost identical. 

2) I am not familiar with SoundRuler, but I checked the manual to SoundRuler. More generally, Correct measurments depend type of the call you are measuring, quality of recording and settings of analysis (spectrogram). This is difficult to judge without seeing your calls. With this in mind it seems that it is as easy as you describe in SoundRuler. Apart from the values, you should also see your measurments represented in the call.  This allows you to judge whether the software does what you think it does...

Measuring maximum amplitude does not make sense in most of the cases as it depends on recording conditions (e.g. distance to the animal). You would need to calibrate your recordings and record from the same distance. This is usually very hard in field conditions.

Good luck!

Hi Laura,

I agree with Pavel regarding the channels and amplitude. I also don't use SoundRuler (sorry!) but I thought it might be useful to add that you need to be sure there is no background noise overlapping your calls of interest. If for example these recordings were made at a zoo, there may be visitors chatting, or in the field there could be other animals calling etc. on your recordings. If there is, then you can either need to filter it (if it does not overlap in frequency with the monkey calls), or if that's not possible, you could manually extract the frequency measures, or simply eliminate those calls from your analyses. It might be that you can use the read-out on all your "clean" calls (i.e. no background noise, one individual calling at a time) and in that case it could be just as simple as you say!

Good luck with your interesting project!

All the best,

Esther

 Hi Dr. Linhart and Dr. Clarke,

I really appreciate your answers! I understand what you say about amplitude, I hadn't thought about that previously. My study was in the field and I don't know the distances to animals, so I'm not going to use this parameter. Instead of it, I can measure the 'maximum frequency' (the loudest frequency, which is also the peak frequency). 

Regarding channels and background noise, I've changed my software and I'm now using Audacity. It's easier for me and I can better eliminate the noise. In this case, I can separate the channels and only analyse one of them, as you said. But I've tried to measure both and as Pavel said they seem to be identical, so this is not a problem anymore :)

Thank you again for your thoughtful answers! :)

Laura

Hello Laura.

I used SoundRuler in the past. Great software but not sure it fits with what you plan to do. My suggestion is to use Praat for your analyses and follow the suggestions you have already received about channels and amplitude. Best wishes. Marco

Thanks Dr Gamba!

Yes, I realised that SoundRuler doesn't fit exactly with what I want, so some days ago I started using Audacity and I'm making strides in my study! But thank you anyway, I'll take Praat into account for the next time I get in trouble :)

Laura

Hello Laura,

I will probably repeat some aspect of the other answers...

I also use Audacity, but only for basic editing. You mentioned it is a good software for dealing with noise, but I don't recommend using this feature in samples that will be analyzed. The problems are (1) it is destructive, so be sure to keep backups of the original recordings, (2) those noise removal methods generally do it in an adaptive way, so that signals embedded in different noise levels could be affected differently; (3) in Audacity those methods are essentially aesthetic, if noise removal is a critical step for you, look for other software. In short, if you want to get standardized measurements don't do noise removal at all. One way out of this is to keep in your data set only those above a given signal to noise threshold (as low as possible). Zoom H4N is very noisy, so you will possibly be throwing away a considerable amount of recordings if did it in the studied species habitat. I strongly advise not to force low-quality samples into your analysis, if you want to produce relevant results.

Good luck with that!