Hi everyone,
I have a pipe section and i'm trying to extract its features using acoustics. Employing a series of Gauss audio files and a sweep file as outgoing signals. Echo data is recorded using microphones near the speaker.
The Gauss signal is very good for evaluating the system using cross-correlation technique. However, I'm also trying to evaluate impulse response (IR) of the system using the sweep data. I've written some Matlab codes, but I'm not certain of the results.
Is there some sort of open source code (calibrated) or some software that I can use to compare/contrast my results against?
More system details:
Empty pipe system, contains bends, junctions, holes, and diameter changes (variable). Audio frequency range. No clipping or saturation observed.
What you are trying to do has a lot in common with channel estimation using blind deconvolution. There is a huge literature out there on these techniques. My area is electromagnetic wave propagation, but I imagine that much work has also been done in acoustics, particularly in seismic sensing, room acoustics, and acoustic defect detection in solid materials.
You should be able to verify your results experimentally using pipe sections whose impulse response is known ( where modes of vibration can be analytically determined).
Thanks George. The techniques are similar between EM and acoustics (with acoustics applications in room acoustics, ocean acoustics etc).
I have lots of experimental data investigating pipes buried underground and trying to extract the features acoustically. I have code that evaluates the impulse response, but I was looking for some alternate technique to compare/contrast my results to improve confidence.
I'm not doing modal analysis as of yet. Just extracting info from echo data.
I believe I understood your problem, so if you're using MATLAB there is a very straight-forward manner of doing this. For sweep signals, you can always rely on basic signals theory. For an input x(t) and an output y(t) affected by system h(t),
X(jw) * H(jw) = Y(jw)
H(jw) = Y(jw)/X(jw)
So all you need to do really is to divide point-by-point the vectors of the FFT from the input signal and the output signal. In MATLAB, download and setup the ITA-Toolbox, from the folks at Aachen, and do this
x = itaAudio;
y = itaAudio;
x.timeData = %(here comes the data from the input signal)
y.timeData = %(here comes the data from the output signal)
h = y/x;
And, as simple as that, you have the system impulse response in the 'h' variable.
HOWEVER,
Since you're dealing with tubes, I'm afraid that results such as Transmission Loss won't be anywhere near what you'll find by this method, since there are reflections and, at about any point in the tubes you're measuring, there are "forward travelling" and "backwards travelling" wavefronts. To measure such parameters correctly, therefore, you would need to use a more robust technique in this sense. I personally would recommend trying out the four-load method since it provides very reliable measurements.
I hope your measurements go well ;)
- Badges
- Science topic
- Similar topics
- Filing