Question
Asked 6th Jun, 2014

Is it possible to estimate SPL (dB) of bird calls from .wav file (audio recording)?

I am working on avian acoustic adaptations in urban habitats. I use Avisoft SAS LabPro and Raven Pro1.4 software. These software do give amplitude levels of sound but are exceptionally too high in comparison to measurement made by Sound Level Meter SVAN 957. What measurement setting should I do in the software for getting accurate SPL( dB) from the .wav files?

Most recent answer

2nd Feb, 2018
Manoj Singh
Jawaharlal Nehru University
thanks Sachin for the answer

Popular Answers (1)

2nd Feb, 2015
Russ Charif
Cornell University
Your question requires a bit of clarification. When you talk about estimating SPL of bird calls, do you mean the received level (RL) of the calls at your microphone, or the source level (SL) as the bird produced the sound? SLs are typically given in dB SPL (or dB re 20 µPa) at a distance of 1 m from a source.
If you want SLs you first need to measure (or estimate) the RL and then apply an adjustment for the distance between the bird and your microphone, since RL falls off logarithmically with distance from the source.
In order to get actual RLs from a recording, you need to have calibrated (or more properly “characterized”) recording equipment so that you know the scaling relationship between the digital sample values in the audio recording and actual sound pressure at the microphone, measured in micropascals (µPa). There are two approaches commonly used to obtain this calibration info. The first “stepwise” calibration approach relies on knowing the way in which the signal is transformed at three stages: (1) the microphone, which converts pressure variations into voltage variations, (2) the preamplifier, which amplifies the weak electrical signal from the mic, and (3) the digitizer (analog-to-digital converter) which converts the continuous voltage waveform into a series of digital audio samples (typically 16- or 24-bit integers).
The second approach is to treat the entire recording system (mic + preamp + digitizer) as a “black box” and just determine what the scaling relationship is between the audio sample values and the sound pressure (in µPa) at the microphone. Usually this is done by making a recording of one or more calibration signals (usually pure tones) while also measuring the level at the microphone with a sound level meter.
These approaches are summarized in the attached Powerpoint slides.
Beginning in Raven Pro 1.5 (currently in a beta version) you can enter calibration information if you have it available to get true SPL values out from the measurements. Without that calibration info, dB measurements in Raven can be used for making relative comparisons between sounds in the same recording (or in different recordings if they were made with the same equipment and gain settings), but not to infer actual dB SPL.
One caveat to keep in mind is that if you change gain settings during a recording or between recordings, you need to know what the gain setting was at every time, and what this gain setting corresponds to in terms of dB of amplification.
3 Recommendations

All Answers (8)

6th Jun, 2014
Nike Gnanateja
University of Pittsburgh
Hi manoj,
The sound processing softwares commonly give the amplitude values in dBu (Arbitrary units) which is not equivalent to dBSPL. Measuring dBSPLs can be reliably done only by using calibrated hardware (sensitive microphones with sound cards) with its associated software. This job is best done by the SLM. The tech notes of the SLM you mentioned say that it has option to record sound, why don't you use that option for recording and estimating SPLs and do any further processing in your softwares. (a more elegant idea would be to route the sound from the SLM to your software, if your SLM has an audio out.
If you still want to estimate the SPLs using your softwares there is a workaround but is a little intensive (though I dont recommend this). You have to derive a correction factor for converting the arbitrary units into dBSPL. You can do this by recording sounds (prefereably pure tones) of different intensities ( at least 10 steps) using both your SLM and the recording device with both the microphones placed together. Then tabulate the readings of the softwares, run a regression analysis with the SLM reading as the dependent and the software reading as independent variable. Then use the regression weights as the correction factor. You can also fit a logarithmic function to the data using the cftool in Matlab and derive the correction factors.
P.S: I wonder if the distance between the bird's mouth and the microphone is constant!!!!
6th Jun, 2014
Jernej Polajnar
National Institute of Biology - Nacionalni inštitut za biologijo
It is also important to note that your conversion ratio might be variable if you used more than one device for recording. It can also be something as banal as inadvertedly changing the microphone input volume in Windows on the laptop that served as a recording device (for example if more people shared the machine and someone used it for skyping when you weren't in the field). Therefore, I wouldn't advise relying on specific values if you aren't completely sure that the input volume didn't change.
9th Sep, 2014
Carlos B. Araújo
Universidade Estadual de Londrina
You can do that if you calibrate your recorder using a sound source of known Sound Intensity Level. Still, you have to understand a little the differences from Leq, RMS and Peak.
2nd Feb, 2015
Russ Charif
Cornell University
Your question requires a bit of clarification. When you talk about estimating SPL of bird calls, do you mean the received level (RL) of the calls at your microphone, or the source level (SL) as the bird produced the sound? SLs are typically given in dB SPL (or dB re 20 µPa) at a distance of 1 m from a source.
If you want SLs you first need to measure (or estimate) the RL and then apply an adjustment for the distance between the bird and your microphone, since RL falls off logarithmically with distance from the source.
In order to get actual RLs from a recording, you need to have calibrated (or more properly “characterized”) recording equipment so that you know the scaling relationship between the digital sample values in the audio recording and actual sound pressure at the microphone, measured in micropascals (µPa). There are two approaches commonly used to obtain this calibration info. The first “stepwise” calibration approach relies on knowing the way in which the signal is transformed at three stages: (1) the microphone, which converts pressure variations into voltage variations, (2) the preamplifier, which amplifies the weak electrical signal from the mic, and (3) the digitizer (analog-to-digital converter) which converts the continuous voltage waveform into a series of digital audio samples (typically 16- or 24-bit integers).
The second approach is to treat the entire recording system (mic + preamp + digitizer) as a “black box” and just determine what the scaling relationship is between the audio sample values and the sound pressure (in µPa) at the microphone. Usually this is done by making a recording of one or more calibration signals (usually pure tones) while also measuring the level at the microphone with a sound level meter.
These approaches are summarized in the attached Powerpoint slides.
Beginning in Raven Pro 1.5 (currently in a beta version) you can enter calibration information if you have it available to get true SPL values out from the measurements. Without that calibration info, dB measurements in Raven can be used for making relative comparisons between sounds in the same recording (or in different recordings if they were made with the same equipment and gain settings), but not to infer actual dB SPL.
One caveat to keep in mind is that if you change gain settings during a recording or between recordings, you need to know what the gain setting was at every time, and what this gain setting corresponds to in terms of dB of amplification.
3 Recommendations
2nd Feb, 2015
Gianni Pavan
University of Pavia
if you use 1/4" omnidirectional microphones you can use a sound calibrator that fits on the microphone capsule. They usually produce 94dBSPL at 1kHz. Record the calibration tone at the beginning of each recording, or at least after any level adjustment. This way you always have a reference tone in your recordings to make measures of Received Level. Then you can assess Source Level if you have a good estimate of the distance form the source. The method is great if you have suitable microphones; it works perfectly with autonomous recorders designed to capture sounds at 360°, e.g. for soundscape recording.
However the method can't apply if you use generic microphones, cardioids, supercardioids, or parabolas. In such a case you can use a speaker with a fixed tone; in open field (not in a room!) you can set your microphone and recorder, and use to a Sound Level Meter to measure the level you receive at the microphone and take a note of the level you're recording. Don't expect great accuracy as sound propagation is often complex and received levels may vary. Once you have a good measure for a given setup and recording level ..... don't make changes and do your recording .... If you can lock your configuration (microphone, cable, recording gain and recording level) you can go on for a while ..... but periodically it is recommended to verify the calibration.
2 Recommendations
4th Apr, 2016
Donatas Bagočius
Klaipeda University
I use Raven Pro 1.4. It seems that Sound Level Meter SVAN 957 uses different type of  integration of values in comparison with Raven. Actually Raven even do not have a function of estimation of SPL. Though its posibile to extract samples of power spectra in decibels and integrate them to get SPL. However the Raven has an option of different windows like hanning, rectangular, Kaiser and 3 dB window option with different width. So every time you change these options you will get different result. Even the result with the same calibration file can be up to 50 dB. So the smartest way is to use one software for all measurements and see the change of levels as a Trend with reference to known sound level or chosen one.
2nd Feb, 2018
Sachin Singh
National Institute of Technology Delhi
we can measure with the help of programs in matlab.
Can you help by adding an answer?

Related Publications

Article
Full-text available
This study aims to uncover patterns of species identification error in bioacoustic surveys of central Amazon birds. To quantify errors, we developed an on-line quiz based on vocalizations of an undisclosed set of 41 antbird (Thamnophilidae) and woodcreeper (Dendrocolaptinae) species. We invited experts to answer the quiz and obtained 820 answers fr...
Article
Canary. The Cornell Bioacoustics Workstation. Version 1.1. Canary 1.1 software for Apple computers and the Canary User's Manual Copyright © 1993, Cornell Laboratory of Ornithology. All right reserved. Price US $ 200.00.
Article
Song ontogeny in birds has been studied primarily with estrildids, because these species can be easily kept and bred in captivity. Many field studies have show that song is learned or modified at a later age. Song structures are often adapted so that they match those of conspecifics with neighbouring territories. But, there is controversy surroundi...