Sound Analysis - Science topic

There is no need to include end correction in general in numerical acoustics, as the Finite Element setup does not need to be corrected. The correction is needed for models that approximate the acoustic behavior, such as Narrow Region Acoustics model that approximate exact thermoviscous behavior, or the Interior Perforated Plate which again uses an analytical expression to approximate exact behavior.

SpeechRecorder by LMU Munich University. I would still recommend PRAAT, however. It's simple and there are many tutorials about how to use it on YouTube.

Your understanding of music goes back to the theory of mimesis. The problem of interpreting the phenomenon of music depends on where you draw the line between art and reality.

An important, but somewhat distracting, observation: if one were to find a genetic predisposition to music in some people and not in others, such a theory could become the basis of Nazism, racism, xenophobia in all its possible variations, and the corresponding potential for discrimination.

I use Audiosculpt from IRCAM. https://forum.ircam.fr/projects/detail/audiosculpt/

It produces a spectrogram for audio - with control of the analysis parameters. For data output, you can choose how many partials you want to consider for each sound. It also has tools for fundamental estimation etc.

NB For analysing the numerical data, you need to use an extra small program and then view with a spreadsheet etc.

Thank you very much sir. René Christensen

Spectrogram presented not only the frequency content of the signal but also its energy. The spectrogram's vertical axis represents frequency, with the lowest frequencies at the bottom and the highest frequencies at the top, while the horizontal axis represents time; it runs from left to right of the axis. The colors enrich the spectrogram representation as its third dimension; different colors represent different energy levels . so if you combine it with classifier such as CNN which is manly image classifier it might give good results

Avisoft SasLab Pro 5.2 software has got inbuilt lowpass, highpass, notchpass and bandpass filters critical in elimination of unwanted noise. Open the Avisoft SasLab Pro 5.2 software, the choose the edit menu. Under edit menu select Filter. Under filter, choose the Time domain IIR or FIR Filter. If its the background noise, then you may record the room tone and filter it out. Similarly the Avisoft UltraSound Gate allows you to attenuate unwanted signals.

Analysis starts from technical suitability. Out of 3 companies quoted, go through the specifications and engineering design with Durability. Then u go to the lowest offer suitable to tech spec

CATT acoustic is better for theater simulations in my experience. You may also try to use AcMus developed by a Brazilian team.

Hi Thomas Luypaert , The window size in a FFT analysis represents a number of samples, and a duration. To set the size you would need to look at the fundamental frequency in your system, its intensity and any changes it may experience in time. The window size expressed in samples along with your sampling rate (samples/second) would give you information to calculate the window duration.

Also, from the spectraplus page: " The frequency resolution of each spectral line is equal to the Sampling Rate divided by the FFT size. For instance, if the FFT size is 1024 and the Sampling Rate is 8192, the resolution of each spectral line will be: 8192 / 1024 = 8 Hz. Larger FFT sizes provide higher spectral resolution but take longer to compute."

Hope this helps!

Analyzing a sound's spectral makeup (fundamental frequency plus harmonics) is easy with commercial or free software (I often use SpectraPLUS, which has a 30-day trial). What isn't so easy to access is a sound's intensity or sound pressure level (SPL) unless the recording comes with a calibration file. Inexpensive sound level meters (SLMs) may help when making live measurements, but these SLMs often present sounds as A-weighted only, meaning the measurements filter out very low and high frequencies. Very low level sounds (i.e. very soft sounds) are not easy to measure, as microphone noise alone can obscure the measurement. So, depending on what attributes of a noise or sound you're investigating, publicly available sound snippets may not be sufficient. Eric C. (former Principle Engineer, Doppler Labs; former Chief Engineer, Countryman Associates).

Natural frequencies are almost independent of the tool rotation, what experienced milling operators do is to adapt the milling parameters in a way to avoid excitating natural frequencies hence minimise the tool/piece vibration and get the smoothest result.

Natural frequencies depend of the piece fixed on the milling table, the course of the various X Y and Z slides etc so cannot be determined once for ever for a given milling machine

Devices as "tuned vibration absorbers" can be fixed on the milling head to fix problems that cannot be sorted by adjusting the milling parameters off a risk of resonance, but this requires a solid vibration engineering culture

Dear Sumit,

Sound of cracking crystals between two hard surface depends on couple of important metallurgical as well as mechanics factors such as increasing strains, grain distribution reaches log normal packing, coarsening of crystal grains, mis orientation angles, level of temperature distribution due to fraction between two surface with grain misalignment effect , critical grain size micro-cracking and thermal enhancement, cohesion crack nucleation, defects, cavities etc.

In general, If the size of the crystal grains is closed to the wavelength of the supersonic ray, the part will not be transparent to ultrasound, in this case the there will be no ultrasound bottom of signal. it is also seen that size of the crystal grain increases the intensity of supersonic wave decreases.

Sound products can be listen at high extrusion speeds.

Atomistic modeling is highly applicable in easy understanding of crack sound between hard metal.

Ashish

Thank you. I was able to synthesize them with the software.

Dear concern,

dB is used to quantify the ratio of two values in logarithmic scale which conveniently represent very large or small numbers on a same scale.

For converting the ratio of two power values in dB, we use ans(dB) = 10*log10(ratio) and ratio=10(ans(dB)/10).

For converting the ratio of two voltage or current values in dB, we use ans(dB)=20*log10(ratio) and ratio=10(ans(dB)/20).

You can verify the concept of multiplicative factor of 20 and 10 using P = V2/R = I2*R and taking R=1.

For your information, dBm is used to quantify the value of Power (not a ratio) in logarithmic scale which is generally used in wireless communication and other areas.

Hope you find this answer hepful.

Unqualified comments :-

Time weighted negotiating of a mass, collection of matter considering field effects ( say gravity, possibly friction and so on) with a directional effort is work, forced displacment sort of effects.

Rate of doign work is power (watts) , sort of time rated expense of energy. Describing power has nothing to do with underlying phenomenon.

Most important thing to understand or find or reconcile is

proxy of or displacments/field

proxy of or efforts,

and of course ratio of expended against recoved power offers clues about nature of losses , mostly thermal, and proverbial efficiency etc.

Dear Tatiana Izmaylova . The missing fundamental is "missed" because it does not appear in the frequency analysis performed by a Fourier Transform or another similar technique. What is present in the original signal is a number of harmonics that corresponds with a fundamental frequency fi=n*f0. And this is the "physics" of the phenomenon, i.e., the reality that is measurable.

If the fundamental frequency were present in the signal, we could measure it with the appropriate techniques. But the acoustic system is unable to generate this particular frequency because of its size or other physical characteristics.

It is when our brain interprets the received signal, in terms of pitch, when we can describe the sound by means of a tone that should have a lower measured frequency.

To obtain a fundamental tone of f0 from a modulation of, let say, a second harmonic 2f0, we will need a modulating frequency of f0 (the result of an amplitude modulation of two tones are the sum and the difference of the modulating frequencies). With the appropriate bandwidth in the frequency analysis, we should be able to separate the two tones. A long enough Fourier Transform gives us a short enough bandwidth analysis.

For all the reasoning above I think that it is the interpretation that our brain makes, by its previous experience of learning about the pitch of a sound, which produces the phenomenon called missing fundamental.

I agree that Audacity is of no use-- it's primarily an editing program, not for measurement. Praat and SAP would both probably work for the simple measurements you described. I believe Praat (which was developed for work on speech) has some tools for measuring speech-related features such as formants, which might be useful for nonhuman primate vocalizations. Raven Pro (http://ravensoundsoftware.com/) has (I'm told) an easier learning curve than Praat and includes basic time, frequency, and relative power measurements (though no formant measurements). There's also a set of freely available tutorial videos available (see "Training" on that website).

There is a highly instructive user manual on the site. The reference for the initial publication is:

Tchernichovski, O., Nottebohm, F., Ho, C. E., Pesaran, B., & Mitra, P. P. (2000). A procedure for an automated measurement of song similarity. Animal behaviour, 59(6), 1167-1176.

I have studied with Olga Fehér who was herself a PhD student of the first author on the paper. I have learned a few things the hard way. First off, the initial setting of this software is designed for the analysis of zebra finch song. If you are working with a different species you may need to adjust them. Also, the detection and classification algorithms are good, but by no means perfect. Therefore I strongly advise that you keep the results by eye!

Greetings from the UK!

Tarandeep

For most analyses, the biggest problem with clipped recordings is not the loss of reliable information about relative amplitude ("dynamics"); it's the corruption of the spectral content of the recording. Clipping introduces spurious energy into the recording at frequencies that are harmonics of the true signal. In some cases, sounds that are purely tonal (no harmonic content) will end up having strong harmonics in the clipped recording. Sounds that truly do have harmonics (like many parrot vocalizations) will be recorded with distorted relative powers in different harmonics. Most kinds of analysis that aim to classify sounds or measure their relative similarity (e.g., spectrogram cross-correlation, dynamic time warping, etc.) may be severely affected by such distortions.

What about "fixing" a clipped recording? Many digital audio editing programs (e.g., Audacity, Audition) include tools for "fixing" clipping. However, these tools are not intended to provide a perfect reconstruction of what the unclipped signal would have looked like. They're meant for rendering the harmonic distortions of brief periods of clipping in music or speech recordings inaudible to the human ear. This is very different from what would be needed for a quantitative measurement and classification analysis.

Although there may be some kinds of analysis that could tolerate clipping (e.g., analyses focused on temporal patterning of sounds that don't measure spectral content), in general I recommend against using clipped recordings for pattern analysis and classification. You're likely to get quite different results from what you would obtain using unclipped recordings of the same signals.

Hola Miguel Alonso-Cambrón ,

He ojeado la página y me parece un proyecto interesantísimo! Muchísimas gracias por el dato. Me pondré en contacto con Juanjo Palacios para saber más y enriquecer el estudio que quiero desarrollar.

Gracias, saludos!

I think it is possible since even a cat's purring can aid in the healing of injured tissue

If you want to quantify spectral profile by statistical metric and your desired profile is narrow with tall peak.

you can do this by window whose width should be more than desired feature. Calculate the first and second order moments that is sufficient to differentiate narrow and wide features.

You can do this with non-over lapping and over lapping windows.

I think this will help you.

Dear Prashanth,

welcome,

You can use body area wireless system to acquire the sound and send it wirelessly to your computer where you have the identification and supervision system. This may be considered an application for internet of things.

The solution to use a sensitive microphone and amplifier may fail because as the distance increases the noise picked up by the microphone may be too high and impair the sound detection. However it may be a possible solution.

I expect the first solution is the best one.

Best wishes

Hi, i'm an acoustics professor in Colombia and i'm really interested in the project and in general in Master Acoustics International , could you give me more information.

Best regards

Dear Mr.James Garry

It is just a personal interest,my major study is in a different field.

Muchas gracias Gian, you work is very interesting!

Update; I buy an used Olympus LS-100, after reading helpful reviews in birds pages and bioacoustics forums in general. And my microphone is a shotgun Sennheiser (the pic is attached to this reply, i don't know the model). Thanks to all!

Dear Mike,

Xbat is indeed a nice piece of software that is extensible and well-architected although it is no longer supported and I don't know how well it will work with modern versions of Matlab.  Ludwig's other suggestions are fine as well, and I would add Cornell's Raven although it is not free (nor is Audition).  In addition, many people use Praat which was designed for analyzing human speech and has some very nice features although it is perhaps not quite as user-friendly as some of the other offerings.    There are a number of programs designed for analyzing marine mammal calls that could also be useful to you:  Dave Mellinger's Ishmael (http://www.bioacoustics.us/ishmael.html) and Douglas Gillespie's PAMGUARD (https://www.pamguard.org).  Our group also has an extensible Matlab-based analysis program (Triton, http://cetus.ucsd.edu/technologies_Software.html).

With respect to time-frequency analysis (e.g. a spectrogram), you'll need to think about whether or not you need absolute or relative received level.  If you want absolute received level, you will need to adjust for your acquisition system's calibration curve. If you would like to have this done automatically, be sure to investigate whether or not the software package you select can support this.

Best of luck - Marie

Hi Milan,

What is your sound reproduction system? When do you say 60 dB is SPL?

If your reproduction system is headphones, you can use a binaural head to measure the sound pressure level at each ear and calibrate your system. In the case of loudspeakers, I would measure the SPL at the listener position using a sound level meter.

Best regards,

Diego

Thank you Ignacio and Juan David for your answers! :)

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

B&K 2250 or 2270 

Sorry if I wasn't clear enough. On one hand, what I was saying (which is pretty much what Amaya said again at the beginning of her answer), is that using only sound pressure level data, it is not possible to reconstruct the original sound signal.

On the other hand, if you extract the h(t) (impulse response) using the LMS technique, what you get is the response of the system in the frequency domain, there is no original audio per se, you would have to create a signal based on the response of the system, frequency components, relative amplitude, and what is also important, phase information that is not present in the frequency domain. The result would be very similar to the sound produced by the horn, differences will be observed accordingly to the excitation used in order to produce sound with it.

Regards

Thank you everyone, it really helps! I will soon let you know my choice and how does it sound.

Please see this paper and its references:

If I remember my SPSS output, each column will be what you muiltiply the variable values for (check in whatever you fav SPSS book is, my recommendation is Field's Discovering Statistics with SPSS). THis will give you values (the linear combinations of the variables) for each of the three components and you can use these to calculate component scores for each case in the original datafile.  You could then plot these three new variables and compare them for the two groups.

Hello,

You could use the R packages seewave, tuneR, soundecology and ineq (https://www.r-project.org/). “R” is free and useful in many ways. Plus, there are numerus articles using these R packages for bioacoustic and ecoacoustic research.   

Nevertheless, for a more basic and user friendly approach, you could use Audacity (http://www.audacityteam.org/).    

Best regards,

Aggelos

The Bird Audio Detection project may be of interest, although the aim is detection rather than identification. The dataset has 17000 recordings tagged as either containing or not containing birds. The survey paper reviews the most common methodologies.

If :played back "means sounding all at once in the same place, your question devolves mathematically into asking how many numbers can sum to, say, 3.  Mathematicians are inclined to invoke such ideas as infinity in such cases, though the answer is clearly larger than we can name.

If you at what point does adding yet another sound to a large mix become irrelevant, the answer is also clearly large, but far from infinite.  Having sung in large choruses, I opine that after about a hundred or two voice, the addition (or subtraction) one more (or less) hardly matters in any musical sense.  (The gospel choir at UCSD once numbered over 500, but their major problem was finding a space in which they could rehearse).

The interesting part of your question lies in between 0 and infinity.  Modern psycholofy and cognitive science both suggest that  7 plus or minus 2 is about the limit of our "attention".  But they also suggest we don't multitask, but scan, i.e., shift our focused attention rapidly among competing messages.  Any instrument-rated pilot can tell you that simultaneously scanning (meaning observing and understanding) 7 or so instruments (as well as looking out the window) stretches human capabilities.  Not impossible, but difficult, especially if your life depends on it.

So even though I can't truly answer your question, my best guess would be the well-established psychological number associated with short-term memory: seven plus or minus two.  Seems our brain - evolved over million of years - have figured that before that number, we stand a chance to figure things out.  After that number, run!

As you focus on sound analysis, you may find interesting nowadays state-of-the-art technique for improving speech recognition acoustic models by augmenting training data using with speed, frequency and tempo warping, which is indirectly related to synthetic data generation ( http://www.danielpovey.com/files/2015_interspeech_augmentation.pdf )

Another example of creating synthetic data is block mixing that seems to be useful in polyphonic acoustic event detection: https://arxiv.org/pdf/1604.00861.pdf

We also tried the latter in recent DCASE challenge gaining a little bit on F-measure, but not as large as the author of previous work: https://www.researchgate.net/publication/306118891_DCASE_2016_Sound_Event_Detection_System_Based_on_Convolutional_Neural_Network

Hi Bin, you might want to look at the research leaders in the field of applied soundscape. 

You can look at this book for more details: https://www.crcpress.com/Soundscape-and-the-Built-Environment/Kang-Schulte-Fortkamp/p/book/9781482226317

Regarding your application, you might want to look into the work of Dr. Pyoung Jik Lee: https://scholar.google.com/citations?user=TYZB3ZUAAAAJ&hl=en

We use signal-to-mask ratio in digital signal processing as a measure of how "loud" the noise is compared to the masker. I believe Dr. Lee's work has a range of -3dB to +3dB. for signal-to-mask ratio. I can't be sure, let me know if you find any discrepancies. Hope this helps.

There are two excellent alternatives (free software):

For R users, the package "seewave": http://rug.mnhn.fr/seewave/

The Brazilian biologist Marcos Gridi Papp developed Sound Ruler: http://soundruler.sourceforge.net/main/

Both are excellent options.

Of course there are commercial software (most of them with free "light" versions)

Best

Sorry for the delay in reply Sir and ma'am. Thank you for responding..

I do realize that my question was poorly formed. I am still not too clear about what I am asking , but I have a strong feeling that what i am asking is not too wrong. And am not even sure if I have a clear question, rather it turns out to be a thought. But, would be great if I can get comments or inputs

The Low SR fibres being affected in NIHL was the primary motivation of this line of thinking.

Vanaja mam, asked a very important question. This was one major concern which I was thinking of as well. Rather I was thinking on two lines.

1) when a modulated puretone reaches a receptor, then the the sound varies in intensity from high to low and keeps alternating. Now is the high intensity portion preferentially coded by the low SR fibres and the low intensity portion by the high SR fibres?  In this case, shouldn't modulation detection with affected low SR fibres perception and coding of shallow modulations should be affected than deep modulation. Also by extension modulations at high SLs should be more affected than modulation at low SLs

2) When two very close frequencies (within an auditory filter) are presented one at low and one at high intensity. Then the low intensity signal will fall within the excitation pattern of the high intensity. How are the different SR fibres coding the signal now. We also have to consider the weird phase relationships that may be present between the two signals.

Please do comment about these. I don't have a very specific question here though, but await a good discussion.

best

Nike

Hi Jonas, thank you for your answer. It certainly is close to the topic. I think I found what I was looking for. Actually, it was a study by one of the researchers here at the IHR: 

Briley PM, Krumbholz K. The specificity of stimulus-specific adaptation in human auditory cortex increases with repeated exposure to the adapting stimulus. J Neurophysiol [Internet]. 2013;110(12):2679–88. Available from: http://www.ncbi.nlm.nih.gov/pubmed/24047909

This study was able to show adaptation as a function of adapter-probe frequency separation. Adaptation is stronger when there is no Δf, but it is still present (although in a lesser degree; 50% adaptation) when the Δf is as high as 1.5 octaves.

Hope this is useful for others.

Kind regards

Nuno

Van Leeuwen's Speech Music Sound may also be relevant to your multi-modal perspective, providing terminologies and various examples.   

As suggested by Israel, if the "nature" of your noise does not change (assuming frequency remains the same as the signal but not the amplitude or phase) then a simple amplitude based filter should work. You may also try phase-based filters.

If the noise (or even if the signal) changes its characteristics then go for statistical filters which can even be adaptive! The first one to try out is Wiener filter mechanism with adaptive filter algorithms like LMS, NLMS and RLS. These are the best ones (my personal experience). See these for more information 

and

and

I recommend you to read a brief introduction to sound analysis, for example:

A good reference book also can be

Au, W. W., & Hastings, M. C. (2008). Principles of marine bioacoustics (pp. 121-174). New York: Springer.

ISO 690

I guess you are trying to analyze your dolphin whistles using contour analysis, there are many publications on this technique, you can check this recent ones and track back to the older ones:

Gannier, A., Fuchs, S., Quèbre, P., & Oswald, J. N. (2010). Performance of a contour-based classification method for whistles of Mediterranean delphinids. Applied Acoustics, 71(11), 1063-1069.

Ferrer-i-Cancho, R., & McCowan, B. (2009). A law of word meaning in dolphin whistle types. Entropy, 11(4), 688-701.

There is no general method to establish a correspondence between the two measurements.

To establish a "correspondence" you will need characterization and modelling tools. There is also conditions on the tested material.

In a Kundt's tube, the acoustic excitation is plane waves at normal incidence thus the TL should rather be called normal Sound Transmission Loss (nSTL).

You also have samples with small dimensions and this will tremendously impact your nSTL at low frequencies (when the wavelength in air is very much larger than these dimensions). Indeed, the nSTL can reach high values at low frequencies as long wavelengths will not be able to propagate through a small sized sample and will be reflected. This is a measurement artifact you will not see (or to a lesser extent) in ISO 140-3 measurements.

When the frequency increases you should find a nSTL following the mass-law behavior.

At even higher frequencies you might find bending vibrations of your homogeneous sample. From the frequency positions of these vibrations you can estimate the elastic parameters of your samples. Of course this is easier if your sample is made of an isotropic material and its first bending frequency is below the highest frequency of plane waves assumption in the Kundt's tube. Honestly, if one of these 2 conditions is not fulfilled, forget the Kundt's tube.

If everything is OK at the previous step, from the mass and elastic parameters of your sample estimated from a Kundt's tube measurement you can estimate the sound transmission loss for a larger sample (~ 10 square meters is required by ISO 140-3) and for a diffuse sound field with a modeling tool.

Finally, you're right, there is no ISO standard for the measurement of the normal Sound Transmission Loss while there is the ASTM E2611. The ISO is aware of this point and a working group should start a discussion about it in the next few months.

According to what you mean by emotion, you may look to Stephen Mc Adams and Albert Bregman's work.

In France there were also interesting works such as even itf they are not exaclty on emotion these artistic, creative and penomenological approachs can help to qulify emotion:

La revue l'Espace du Son

My works on spatial sound  :

ok.

Hi Yopie,

Behringer has a device on which you can tap a button to the beat of the song or .wav file.  And many Pro-tools plug-ins have a "tap tempo" function which then calculates bpm for you.  In the simplest sense, 60 bpm is 1 beat per second.  So, if you are tapping your finger 50 times a minute to your .wav file, the bpm = 50.  This assumes you are using quarter notes when tapping however.  If you tap your finger to your wav file and you noticed you tapped 100 times in a minute, this could mean that you were tapping eighth notes and really, it was 50 bpm and not 100 bpm, since bpm, in the classic sense, is always based on quarter notes. If you are are dealing in different time signatures (e.g. 7/4, 5/4, 3/4, 11/4) you always find what the quarter note pulse is and find out how many times you tap your finger in a minute - and that is your bpm.  Be aware of what a quarter note is - and this should solve it for you. Or, buy a metronome and see what the metronome setting is when it seems to be 'in sync' with your .wav file. What is your .wav file?  A song, sound of ocean waves, something other than music? Heartbeat?  Also, you could have something weird like 90.5 bpm.  As well, if you listen to songs that were not recorded with a metronome, the bpm changes from time to time throughout the whole song. Early Beatles = no metronome (flowing excitement; the beat make sense as the song grows, e.g., "It Won't Be Long"). Sting = perfect metronome (e.g. "Fields of Gold"). Stravinski, Beethoven, Brahms, Mozart... suggested tempos, and rit. (slow downs) and accel. (speed ups)... gas pedal moving up and down. Are you dealing with a perfect bpm?  Only you have the .wav file you are talking about ;]  120 bpm is another very common tempo.  The odd tempos 117, 105, 99 ... have a certain magic to them as well. 

Attached is a tool I made for exploring this sort of thing...

Ole-Herman,

You still haven't told us why you want to use a hydrophone in air.

I agree with everyone on the impedance mismatch. Hydrophones just weren't designed to work in air.

If you are looking for a weatherproof or immersible microphone, there are a number available. Just search the web on "waterproof microphone". Some are fully immersible, although they probably don't work very well as hydrophones, just as hydrophones don't work very well in air.

A weatherproof mic in air will give you a much better signal-to-noise ratio and a much more predictable amplitude-frequency response and directional characteristic than a hydrophone in air.

If you want a transducer that will work in both media, I suggest you use both a hydrophone and a waterproof mic and mix them after the preamps. If you set the gains correctly, the impedance difference between the two media should effectively switch between the two transducers automatically.

In 'Methoden der Mathematiaschen Physik' by Courant and Hilbert is given how you can calculate all modes of a circular plate. The resulting algorithm is easily executed in Matlab.

The MP3 format is a complicated lossy compression format for audio. Writing an MP3 decoder is a major undertaking that requires a fair bit of expertise.

Rather than writing code from scratch to decode MP3 files, you should consider using third-party code, such as the open-source MINIMP3 decoder (http://keyj.emphy.de/minimp3/) or a third-party library like libmpg123 (http://www.mpg123.de/api/). Using such third-party code, you can convert audio data to uncompressed form, which you can then easily export in the format you desire.

[NB: Microsoft Security Essentials flags the .exe file in the MINIMP3 library package as a Trojan. I believe this is a false positive, but just in case, delete the .exe file and focus on the source code, recompiling if necessary using your own system.]

What do you exactly mean by richness? In terms of properties of the sound, richness may be related to the spectrum frequency, to the amplitude evolution through time or, for instance, to the harmonic relation among frequencies (harmonic and partials). What properties are you considering?

I don't know how you approach this topic. But one of the easiest ways would be this:

1. Create/Record the sound.

2. Calculate Spectrum (or Spectrogram if the sound changes over time)

3. Look at the regions with the highest energy. If you have only one maximum it is most likely a monophonic signal.

I was wondering this same question a few years ago. I came to a conclusion that the definition depends on your application. In my field of science which is room acoustics, the very first part of the impulse response, say 5 ms after the direct sound, is considered as signal. Then your SNR is "power of conv(x,h(0:5ms)) / (power of noise)". Thus, the rest of the impulse response, i.e., from 5 ms to infinity is considered as convolutive noise and is not presented in the traditional SNR figure. Instead the convolutive noise is presented by another figure called the Direct-to-reverberant ratio (DRR) and it is defined as "power of h(0:5ms) / power of h(5ms:infinity)". Sometimes, in room acoustics,  also a parameter called the reverberation time is used as the indication of how much convolutive noise there is, as in 

Champagne, B. Bedard, S. ; Stephenne, A "Performance of time-delay estimation in the presence of room reverberation" IEEE Transactions on  Speech and Audio Processing, Volume:4 , Issue: 2 , Pages:148 - 152

That is, the signal-to-noise ratio is defined by what is considered noise and what is considered signal in your application. 

We published a paper on unit and integration testing for the Jamoma framework some years ago: 

A bit more clear. So you have one ultrasonic wave inside the tank - from bottom to top? - and a moving liquid ("wave").

Do you use puls-echo mode or transmission?

Now you want to see the affect of moving liquid in your received ultrasonic signal?

Sometimes a sketch of the setup is worth 10.000 words.      ;-)

I do not think either option is definitely better than another. The key question is how are you going to work on your recordings. If your target reproduction system is a multichannel loudspeaker system, then definitely go for ambisonics, but remember your listener must have his head positioned exactly in the sweet spot to hear correctly.  You will not have a problem of non-individualised HRTF in the recording. The binaural option is an obvious solution if your primary listening tool is headphones, and if you are to be the main listener, then recording binaurally with your own ears seems a better option. If you have both reproduction and recording systems available, then try both and decide which works better for you. If you need to purchase a multichannel system but your budget allows, then also go for ambisonics. A 16 channel system would be fine, inexpensive monitors like Genelec 6010 are OK.  A nearly as good result will be obtained with just 10 speakers: 6 around the listener in the horizontal plane at the level of the ears and four around an elevated horizontal plane.

The US national Park Service has focused a lot of efforts on this Contact Kurt Fristrup (kurt_fristrup@nps.gov)

The most direct and simple method is to take a longer FFT.  This may require zero padding if you do not have enough input samples.  (If the input signal is stable and you have enough data, it is better to increase the FFT size using more input samples than to zero pad.)  This separates the signals into narrower FFT bins and may be enough to separate the two signals.

If that does not work, a slightly more complicated method is to use the Chirp Z Transform (CZT).  If you have access to Matlab or Octave, this is given by the command czt().  Set up the CZT so that it follows a contour along the unit circle in the Z plane (in matlab, parameter A = exp(1i*2*pi*LowF/SR), where LowF is a frequency somewhat lower than the frequency being examined.  Set increments of frequency (deltaF) between points smaller than the width of the FFT bins (W = exp(-1i*2*pi*deltaF)). 

Both of these methods will fail if the signals are too close together and there are not enough input samples.  In this case, more sophisticated methods are available, but from the question it is not obvious that more sophisticated methods are necessary, so I'll start with these simple suggestions. 

Thank you for your idea ,fortunately I solved it.

The lower he density of the glass (plastics for instance), the lower its transmission loss for sound impinging on it, from outdoors,for instance. For a composite window (two or more panes with airspace between) the reduced TL for each pane can be compensated by increasing the airspace distance ("gap"), space permitting.

I suggest you to read "Sound absorption of porous materials – Accuracy of prediction methods" by Oliva and Hongisto http://www.sciencedirect.com/science/article/pii/S0003682X14001261

and also "On the modification of Delany and Bazley fomulae" by Kirby http://www.sciencedirect.com/science/article/pii/S0003682X14001261

You would need a sensitive microphone very close to the chicks but I don't believe you could identify species from the sound alone. Much easier to identify from the videos. If they are Culex, pipiens had a higher frequency than tarsalis when we recorded them in the lab.   Peter

I assume that you are trying to scale different sound files to an approximately equal loudness. If that is the case, you can select a Sound Object in Praat and under "Modify", you can use "Scale intensity" to set a specific dB SPL level (Drop me an email if you need a script that normalizes all files in a directory that way). To determine the actual SPL of the output you can use a sound level meter. 

In general a two microphone method is not accurate. One has to use more than two microphones. For accurate measurements I use 6 microphones. A great advantage of this is that one can obtain an acoustic determination of the speed of sound. We carry out a series of 20 measurements at different frequencies. The unknowns to be determined are 20x2 amplitudes (p+ and p- at 20 frequencies) plus the speed of sound c> So we solve by least square a set of 120 equations with 41 unknowns.  After relative calibration of the microphones carried out at the relevant frequencies (placing all microphones in one plane along the tube near a closed end) we carry out measurements with a closed pipe termination. Typical accuracy is 0.3% in the reflection coefficient. Attached you will find descriptions of typical results of the method.

Note that the impedance tube should be quite rigid! Wall vibrations can dramatically pollute your results.

Note that we use a single frequency excitation and a lock-in post processing method, which is essential to obtain high accuracies. Broad-band measurements are a pain in the neck!

Note that large errors can be induced by the way the porous sample is mounted in the impedance tube. Such errors can be very important for narrow tubes. 

Note that the impedance tube only provides information concerning normal wave incidence. Theory should be used to deduce from those data the absorption under other incidence or in a diffuse acoustic field.

thank you Mr. Elias Randjbaran. Actually, im trying to do impact test on acoustics material. For example sound proof panels, impact or damage peak point till the panel destroy or endurance point. so which test are suitable. tq sir.

Hey

Good question, though I do have a question back. By sound generated by a human agent do you mean that it is generated by and from the human (body, voice, etc.) or just that it is initiated and perhaps modified by a human using external source (instrument, shaking a wind chime as opposed to it being shaken by wind)?

I think both questions are interesting, but I am more curious about the second.

If what you want is do pitch tracking (get a graph that shows you what note is being played at a particular time) you need something like melodyne. All frequency representations mentioned above will give you all the frequency components that are present on the signal, which is what I think you're getting with wavpad.

Hi Mathan,

I suggest you I-simpa, it is an open source sofware for 3D sound propagation modelling  you can down load it at the following link: http://i-simpa.ifsttar.fr/

Kind regards