I have been trying to follow the discussion on filters and what not, w.r.t. to detection of the pingers, but believe the techniques being discussed are way off what you would do to pull such a signal from the noise....
Let me explain:
I work A LOT with aircraft sounds, no need to explain more, but the MOST powerful signal extraction technique I use to isolate specific discrete noise sources from the composite is what I can "profile subtraction" - I suspect there is a "real name" for this but the general process is to sample a long period of baseline noise (for example the cockpit ambient noise in steady cruise, no changing parameters - speed constant, alt constant, etc). This is then 'saved' as the profile. Then I take my wanted sample and "subtract" the profile from the sample... the result is the residual (depending on the FFT depth and bunch of other 'precision' factors), and the quality can be remarkable.
Example: One humorous example came when doing the process described above, when out of the background noise, completely inaudible on the original sample, was a conversation taking place beyond the closed cockpit door, of a rather saucy nature between two FAs... I won't repeat what was said, but the signal must have been 30-35dB below the primary signal level, with the result being perfectly intelligible.
Having once worked on submarine simulators in the past (20 years ago now...) and having been involved in some aspects of the sonar audio, I believe this is a technique that is routinely used.
Anyway, if I were processing the acoustic signals, that is what I would do to pull the pingers out of the background signal.
- GY
