Timeline for Estimating attenuation using sliding windows
Current License: CC BY-SA 3.0
14 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 16, 2012 at 23:25 | comment | added | pichenettes | let us continue this discussion in chat | |
| Aug 16, 2012 at 23:21 | comment | added | pichenettes | Read a bit about cepstral analysis. For speech signals we can indeed take a signal and break it down into a convolution between a spiky signal (vocal cords, "source") and a smooth filter (articulatory system, "filter"). It's because the filter has a smooth shape in the frequency domain, while the source has a spiky shape in the frequency domain. Hence, if you take the spectrum of the log of the spectrum of a speech signal (= cepstrum), the two appear disjoint. | |
| Aug 16, 2012 at 23:21 | comment | added | Nicholas Kinar | Yes, you are right. The original signal is a Maximum Length Sequence (MLS), that has a flat frequency spectrum: en.wikipedia.org/wiki/Maximum_length_sequence. The MLS is bounced off of a reflector, and similar to radar, the cross-correlation of the sent signal and the received signal is the synthetic test signal that I've used. The "switch on transient" is actually (in the model) an initial reflection when the sound wave crosses a boundary between two mediums of dissimilar acoustic impedance. I will show some real data; the real data has a greater bandwidth than the synthetic. | |
| Aug 16, 2012 at 23:11 | comment | added | pichenettes | Could you tell more about the underlying process behind this? Am I right to assume that you send a burst of narrow-band noise (4-20 Hz) - the vertical line at 0.4s being the transient due to switching the source on - into something that affects this burst on the time/frequency? Is it true to say that your "original" signal, even if not known, can be accurately described as a band of noise of constant of amplitude with a switch-on transient? | |
| Aug 16, 2012 at 23:08 | comment | added | Nicholas Kinar | Yes, I think that you are very right about the slow-moving filter kernel and fast-moving pattern of the signal. Can this be used to separate the two? I think that this is also the case with real signals (you are right about the test data being synthetic.) What do you mean about trying one more level of Gabor analysis on the image? I will process some real data to show you and then update my question. | |
| Aug 16, 2012 at 23:00 | comment | added | Nicholas Kinar | Thanks for your response, and for pushing me in the right direction. 1. I might not have access to an "original" signal. 2. Yes, I can make some assumptions about what an attenuated signal looks like. I think that for the attenuated signal, there is a gradual decrease in the power in a frequency band over time. Looking at the original signal, I know that the power of the original signal is somewhat evenly distributed between 0 Hz to 30 Hz. The attenuated signal power gradually "tapers off" toward 0 Hz over time. | |
| Aug 16, 2012 at 22:57 | comment | added | pichenettes | Another naive question: on your image: i.sstatic.net/iRK9u.png, I see a "blue banana". Is that a consequence of your attenuated data being synthetic - obtained by filtering of real data - (on real signals this would be filled by the same "sponge"-like background noise pattern we can see in the upper right corner). Or do we actually see such big dips in the spectrograms of signals in which attenuation is present? The synthetic data you posted is helpful, but it's still not clear to me what your actual input signal(s) is(are) and look(s) like? | |
| Aug 16, 2012 at 22:46 | comment | added | pichenettes | My gut feeling about this would be a 2D flavor of cepstral analysis. You're looking at the product (in the Gabor domain) of two things. Take the log and it becomes a sum. One of them is slow-moving (the filter kernel), the other one is fast-moving (the "sponge"-like pattern of the signal), so they are disjoint in the quefrency domain. So try one more level of Gabor analysis on your images. One corner of the resulting 2D image might be the contribution of your Kernel, another corner might be the contribution of your original signal. | |
| Aug 16, 2012 at 22:37 | comment | added | pichenettes | To turn this into a determined problem, there are two questions worth asking: Do you have access to both an "original" and "attenuated" signal (the problem being akin to system-identification - fitting the ratio of the two to your time-varying kernel model)? Or at least can you make some assumptions about what an attenuated signal looks like? | |
| Aug 16, 2012 at 22:34 | comment | added | pichenettes | In the form you have formulated it, the problem is underdetermined. The reason is that given an input like this: i.sstatic.net/CYPVg.png, what would make me say that it is unattenuated? Wouldn't it be possible that it would be indeed an attenuated version of a signal which originally had a lot of energy in the band affected by the attenuation. It's like taking a signal and asking "recover the impulse response of the filter through which this signal has signal has been filtered". You can't solve that without a prior model of the input signal. | |
| Aug 16, 2012 at 22:21 | comment | added | pichenettes | Thanks for the update, it turns out to be very different from what I thought the problem was - I'm unfamiliar with Sonar stuff in the first place... | |
| Aug 16, 2012 at 18:41 | comment | added | Nicholas Kinar | Thank you very much for your response! Prompted by your response, I've now done more exploring and updated my question above. | |
| Aug 16, 2012 at 8:17 | history | edited | pichenettes | CC BY-SA 3.0 |
added 374 characters in body; added 84 characters in body; added 106 characters in body
|
| Aug 16, 2012 at 8:11 | history | answered | pichenettes | CC BY-SA 3.0 |
