I have continuous recordings of different places in urban and natural areas from which I want to extract the A-weighted background noise level.
Thanks, Francesco. I have the calibration related parameters of the device like microphone sensitivity, gain, AD sensitivity. Is that possible to calculate different noise levels from the recordings?
Akash Jaiswal , I think what Francesco may have meant (please correct me if I'm wrong) is that you need to have recorded a sound level of known level and frequency (e.g. a 1 kHz signal at 94 dB from a calibrator attached to the end of the mic). This value can then be used as a reference against which to calculate the sound levels of your urban/natural area recordings.
I use B&K BKConnect and previously HEAD acoustics ArtemiS. Both are expensive but great tools and the calibration you need relates dBFS to dB SPL
As Francesco Aletta noted, you'll need a reference recording. If you still have access to the equipment used to make the recordings, that should be fairly easy to make as long as you have a way to create a known output level, preferably with an acoustic calibrator.
For open source code, I'd recommend the python-acoustics package: https://github.com/python-acoustics/python-acoustics
It has quite a few different functions which are relatively easy to use if you have Python experience, all referenced back to their relevant ISO standards.
Use A-weighted filter network as output. Then connect to a noise level (statistical) analyzer of the Bruel & Kjaer Type, for example, and sample at 2 seconds interval as the case may be. Noise levels exceeded for different percentages of time and indices will be displayed. I hope this helps!
"As Francesco Aletta Francesco Aletta noted, you'll need a reference recording. If you still have access to the equipment used to make the recordings, that should be fairly easy to make as long as you have a way to create a known output level, preferably with an acoustic calibrator." This is a recommended standard calibration procedure.
Its pretty simple. Use "Poctave" command in Matlab to arrange your data in octave band. And then, subtract the values at each central frequency with their individual weightings.
I have a code for R, but it measures all frequencies, not A-weighted.
Do you still need it?
The wikipedia does have an article on A-veighting giving a regression formula for the A-weighting filter. You need filter frequencies and levels in these filter bands for this to work. Maybe safest to use either 1/3rd octave or 1/1 octave band filtering.
The most important is to have a reference tone on the recording, or being able to determine the reference level. Of course no compression or automatic level control during the recording of the sound files can be used.
It may be beneficial for you to refer annexes associated with ISO717-2. The A-weighting also differs based on the spectrum adaptation terms, i.e. depends on white vs pink noise.
Renata Alquezar Sure. Please, share it. I will explore it. And thanks in advance.
There is a wikipedia article on this with a function which can do the A weighting for you. https://en.wikipedia.org/wiki/A-weighting
Did you record a calibration tone at some point before/after recording anything else? Without such a known reference for amplitude there won't be any way to establish the actual levels. Similarly, the gain settings on the input stage would need to be fixed throughout the recording period.
Yes, Todd Busch. I did record a calibration tone and using it to calibrate my recordings. Thank you.
Regards.
Please note that WAV files potentially incorporate a scaling factor that could change from one recording to the next. As such, even with a known calibration tone as a reference, additional corrections may be needed to arrive at an accurate amplitude.
Todd Busch, I agree. However, my main objective here is to extract information about the noise profile of the sites of my soundscape recordings. I am looking at the variation in the complexity of the soundscapes due to the difference in the ambient noise conditions (anthropogenic). I think, standard calibration will suffice for that though it may not be very accurate for all the recordings. What is your view on it? Any suggestions?
A WAV file can be opened using a text editor to inspect the scaling factor that's been assumed. Alternately, I've seen MATLAB code that extracts the scaling factor in a more automated fashion. You could test out the WAV file option to see if the embedded scaling factor changes from one recording to the next or not. If it doesn't do so for your instrument then you could probably rely on the recordings to be scaled accurately with reference to the calibration tone at, say, 94 dB @ 1000 Hz. If it does vary then you might have to record such a tone at the start of each separate recording in order to ascertain the sound levels accurately.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts