Sunday, April 13, 2014

Automatic detection of MH370 ULB pings

In my earlier post (Analysis of suspected MH370 pings) I manually inspected a audio waveform to determine ping times of arrival. Whilst this is managable for the short segment I looked at, to perform this analysis over many hours or days of recordings would be a bit tedious and probably error prone.

This afternoon I put together a quick MATLAB script to automate this process and ran it over the whole Youtube video audio.

The algorithm is as follows:
  • Read acoustic data from file in 1s segments
  • For each 1s segment use a Hilbert Transform (using FFT) to calculate the envelope of the waveform
  • Calculate noise by smoothing the envelope with a window of length 0.1s (or 10x expected pulse width)
Envelope (red); Smoothed (black)
  • Calculate SNR as the envelope amplitude divided by the noise
  • Locate rising edges in the SNR that exceed the detection threshold (ie start of pulse)
SNR (green) with detected pulse (x)
  • At end of processing plot detection times of arrival module ping interval (1.106s from previous analysis)
Time of Arrival Ping Graph


In the resulting plot that there are 2 distinct sections of coherent detections. The others are false detects from other in-band transients.

Interestingly I had not noticed the 2 individual segments previously, but upon a second listen a distinct discontinuity can be heard hence the 'realignment' of the ping times of arrival. This plot shows that the first segment corresponds to a time where the TPL was going away from the source as the times of arrival are increasing. The second segment has the TPL coming closer to the source near CPA due to the decrease and stabilisation of time of arrival.

10 comments:

  1. This is really interesting analysis. I can only hope the team there in your "backyard" working on the search is as creative and skilled as you are. Thanks for sharing.
    Robert
    Raleigh, NC USA

    ReplyDelete
  2. Dumb question. Can we plot perpendiculars from the segments to suggest the source location?

    ReplyDelete
  3. Thanks Robert! I'm familiar with AJAAC and some of the analysts there. They're a clever bunch and I'm sure they're processing the data they have to extrac the maximum.

    Bruce, spot on mate. That is what I suggest to do at the end of my previous post. I think it is the most accurate way to localise the pinger.

    Even if they only have good CPA data from a single tow track, then the CPA will reduce the along-track error and vastly reduce the potential search area for the CVR/FDR (hopefully the pinder is still attached!)

    ReplyDelete
  4. If the AJAAC uses your methodology, does it seem to you (as it does to me) that if the Bluefin can find it, it is likely to find it sooner rather than later? I.e., because your method seems to suggest a single and precise location?

    ReplyDelete
  5. This comment has been removed by a blog administrator.

    ReplyDelete
  6. Hi Bruce,

    There are limitations to the precision of which you could derive the location of the pinger. The main being the uncertainty in the position of the receiver.

    This position will be estimated based on the depth of the towfish, the length of cable payed out and a model of the drag forces on the cable for the given transit speed and ocean current. I'd expect you could know the position of the receiver to within about 100-200 metres?

    When you add other potential error sources such as acoustic propagation conditions I would be confident you could localise the pinger to within 500m.

    However, this does NOT help the bluefin locate the CVR/FDR as that is not what it is attempting to find. A sidescan sonar operating at 400kHz will be at the limit of its resolution if it were attempting to detect a 30cm wide object.

    It will be looking for larger pieces of debris such as landing gear.

    ReplyDelete
  7. I have been concerned that the search location may have been defined by dubious data refinements from Inmarsat, instead of from the underwater pings but it appears that your confidence in the AJACC analysts is well placed:

    PERTH, Australia— The underwater search of the seabed for debris from Malaysia Airlines Flight 370 is expected to take less time than the previously forecast 6 weeks to 2 months, due to analysis of acoustic data.

    The underwater search area had been significantly narrowed through the detailed acoustic analysis conducted on the four signal detections made by the Towed Pinger Locator on ADV Ocean Shield, Australia’s Joint Agency Coordination Centre, which is leading the multinational search operation, said in a statement.

    ReplyDelete
  8. Rodney, do you have any further comment on the progress of the search and the likelihood of success? If they are running perpendiculars off of four signal detections instead of two, presumably they have more prime search areas to cover. On the other hand, if the perpendiculars intersect at significantly different points, that tends to suggest that all of the signals may be false positives, does it not?

    ReplyDelete
  9. According to the latest JACC media release, "The focused underwater search area is defined as a circle of 10km radius around the second Towed Pinger Locator detection which occurred on 8 April." This sounds to me as though the intersecting perpendicular search, if any, have come up dry so to speak. Does that sound correct?

    ReplyDelete
  10. @Bruce:
    I get the feeling they may not have obtained enough data to localise using this technique. To do so requires pre/post CPA measurements from a single run and if you look at the detection locations (http://www.abc.net.au/news/image/5378120-3x2-700x467.jpg) they only have 1 good long run which MAY encompass a CPA.

    However they can still get an estimate of relative velocity to the source location at each of the pinger reception locations based on tow speed and the resulting ping interval (assuming the transmitted ping interval stayed consisten across the 3 days)

    ReplyDelete