UNKNOWN ARTEFACT: Sound Engineers, apply here!
- Thread starter sohara
- Start date
So... I've been doing a LOT of listening to the Artefact and haven't once heard the same pattern repeated.
Lack of pattern repeat often indicates:
1. The message is very large.
2. The message is encoded with a time stamp.
3. The audio is dynamically generated garbage.
There's definitely a pattern to the audio - there's no denying that. I wonder what 1.3 will bring.
Lack of pattern repeat often indicates:
1. The message is very large.
2. The message is encoded with a time stamp.
3. The audio is dynamically generated garbage.
There's definitely a pattern to the audio - there's no denying that. I wonder what 1.3 will bring.
Q: How many Artefacts are there?
Q: Are the sounds from all of them if there are a few, identical? are there any differences?
I've listened to one I know of on You Tube, It sounds insect like, chittering, almost Snoring! not sure about the howl though.
Sounds similar to the War of the worlds, UuuLLAAAHH! howl.
Looking at it, it looks like a pod attached to an engine device, i.e. exhausts protruding from the rear compartment.
Q: Are the sounds from all of them if there are a few, identical? are there any differences?
I've listened to one I know of on You Tube, It sounds insect like, chittering, almost Snoring! not sure about the howl though.
Sounds similar to the War of the worlds, UuuLLAAAHH! howl.
Looking at it, it looks like a pod attached to an engine device, i.e. exhausts protruding from the rear compartment.
Last edited:
1. I think 4 but one isn't confirmed to my knowledge.
2. hard to tell due to the time limit on listening from the decay of them which can only be stopped by doing the duplication bug so it seems but haven't done that as we would then be exploting. However something noticed in the twitch stream video and video I put up on youtube where we did arftermarth tests on the UA is that the whale sound can only be heard from one UA and both light up when that whale sound happens, you can watch the videos to see if you can get anything better from it.....
Last edited:
Cheers RATKING15, will do.
I looked at the FFTs a while ago and found it difficult to consistently identify the overtones (or base frequency) over the background noise. It would be interesting to have a deeper look though - if nothing else I'd like to know how the frequencies change over time (is the increase from segment to segment, or is it linear across the whole recording?) I've got an old matlab script somewhere that I could dig out that could help with this by giving more control over the FFT. I'll try later on if I get chance.
If a sine of a base tone has been subtracted, then I think it it should be possible to identify where it was and add it back in as the base notes will not be clean sines so I think there will be visible artifacts. I suspect this would just change the timbre of the sound - which might fit with the 'shadow' comment.
Yes, did that. Thought it would be a great idea but since the segments are not of equal length, the result was just garbage. Maybe if you find a consistent marker in the sound to which you could align the segments it would be worth another try but I didn't see any of those and gave up the lead.
I have been wondering if we might be looking at the message structure in the wrong way. Could there be a single message, but we only get to see it in 7-bit segments?
As some purrs fall on the extremes of the envelope and are barely audible, this seems a possibility. If the purrs were all intended to hold information in isolation (a letter or a word for example) I would not expect to see this as it would be important to perceive every bit. To illustrate this, imagine the envelope below which exposing a 7 bit segment from a longer message. With enough segments, it would be possible to reconstruct the original message by identifying points of overlap.
____________/‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾\___________________________________
p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p
So rather than looking for repetition in segments, we should perhaps also be looking for overlapping sequences within the purrs as these could indicate places segments could be joined.
Entirely speculation at the moment, but we should be mindful of all interpretations.
As some purrs fall on the extremes of the envelope and are barely audible, this seems a possibility. If the purrs were all intended to hold information in isolation (a letter or a word for example) I would not expect to see this as it would be important to perceive every bit. To illustrate this, imagine the envelope below which exposing a 7 bit segment from a longer message. With enough segments, it would be possible to reconstruct the original message by identifying points of overlap.
____________/‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾\___________________________________
p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p
So rather than looking for repetition in segments, we should perhaps also be looking for overlapping sequences within the purrs as these could indicate places segments could be joined.
Entirely speculation at the moment, but we should be mindful of all interpretations.
Last edited:
Forget what I said earlier. Was complete rubbish. Removing a sine wave from the recording of an instrument wouldn't remove the note played on that instrument, because the base note is not a sine wave. If it was, it wouldn't have any overtones and would just be a single line in the spectrum. My failure to predict the resulting spectrum made me think about what the result would look like. So I just tried the procedure. I took a recording of a cello playing a C1 (66Hz), loaded it into audacity and saw parallel lines in the spectrum (as expected at 66Hz, 132Hz, 198Hz, etc.). Then I applied a notch filter at these three frequencies. Result (highlighted part is filtered):
You can listen to it at http://www.swedorn.net/cello-notch.mp3. So, maybe someone finds such patterns in the recordings of the UA.
Entirely possible. But why would you assume there will be overlaps in the message parts? Or even worse: some parts of the message ARE garbled (overlapped by other sounds, intentionally, maybe). Still worse: there may be even gaps in the transmission. In such cases the only hope we have is to find common features in the recording to which we could align the parts in order to piece together the message bit by bit.
What I find most interesting is that while whale and tuba sounds do overlap, the chittering is always "clearly" audible, as it always follows the whale sound and seem to never run into the tuba sounds.
I wouldn't assume there are overlaps - Was really just noting that if there are overlaps it could point to an organisation that I don't think has been considered (or at least discussed) yet. Just another place to look for clues. Having overlaps would make it easier to reconstruct the message, and ultimately the message would need to be reconstructed by its intended recipient (whoever that is).
All possible. There are definitely overlapping sounds - I think a lot of the sounds are may by overlaying a number of different samples and those samples change individually over time which introduces really subtle variations. Whether this is relevant though or not is anyone's guess.
Yes, at the moment I'm assuming the chittering is a separator for message segments, the tuba sounds are the main carriers of information and the whale noises are possibly identifiers for interlaced segments.
For me the most interesting feature at the moment is the variation seen at the start of the recordings where the number of tones and the timing of the tuba noises and the timing between the whale sounds is different from the rest of the recording.
Last edited:
Here's my speculation to add to this. I'm by no means an expert in radio or signal analysis and won't be offended by anyone more informed shooting this theory down.
When you demodulate an AM radio signal you pass it through a diode (only lets either the positive or negative half of the wave form through) to filter half of the signal envelope, otherwise the embedded sound is more or less cancelling itself out. There are a few other things that make me think there is an AM signal there. Looks to me like the repeating "white noise" sitting at 1.5 khz (not true AM radio range) but could be an amplitude modulated signal. Looking at the wave forms I think it's 150% modulation, full envelope about 5 times the amplitude of the subsequent envelope in patterns that repeats very consistently. Similar to overtones there can be "sympathetic" EM responses to a carrier signal. Looking at peak levels for an isolated range of 500hz to 2khz there are separate peaks 500hz above and below what might be the carrier signal. Indication that the signal if present could be in the vocal frequency range of a human. Possible indicators of wide band signaling since this high center and 2 side band peak pattern is repeated at 4 frequency elevations evenly distributed through what appears to be the max frequency spectrum for the audio.
I don't have software to attempt to demodulate the signal but I was able to use bandpass filters to isolate the sounds I'd like to experiment more with and can breadboard a simple demodulation circuit. Also have access to an oscilliscope I could look at the audio with...
One thing I can confirm about the UA is it's ability to bring out the OCD in people.
Update: After finally getting some time to investigate this idea further it was a dead end.
Last edited:
Evening gang. Sorry if this has already been asked. I'm a little out of it after a long commute home. Has a anyone done an information theory analysis of the data yet to see if there is even an rational thought being expressed . In information theory if an idea or thought is being expressed then if you arrange the words or variables from least used to most used and plot them on a graph you should get a 45 degree slope. If its just gibberish you will get something close to a horizontal line.
Last edited:
From what i see using Adobe Audition is that it seems to be a broken message. im gonna try and remove the spaces and put the different howls and chirps in the order they appear(while also amplifying the sound) so we might be able to get an audible message. If that doesnt work ill change up the speed it is played and run it every way possible(2x speed,4x,8x,reverse etc.). Also i may just randomly just arrange the segments.
B.T.W im no where near a sound engineer....
B.T.W im no where near a sound engineer....
Last edited:
Correct ... sort of ... I remember it a bit differently but the first order approximation of the number of occurences of all symbols will deviate from a horizontal line in the presence of information iff the sender and receiver of a signal agree on a common protocol (at least know the symbol boundaries in the signal sent). And the latter part is the problem in this case. No one knows which sounds might be the symbols which could form a hidden message.
I was looking at the chittering sounds again but the spectral plot was just too fuzzy to see anything. So I took one recording and did an ICA with the JADE and RADICAL algorithms in octave. Results were nearly identical but JADE was about 20x faster. As expected, the seperation was not that good (with only two sources) but it did clear up the spectrum and I think I found a new type of chittering (highlighted):
You can download the audio here. For the spectrum I used a Hanning window of size 4096. Not sure how to proceed ...
You can download the audio here. For the spectrum I used a Hanning window of size 4096. Not sure how to proceed ...
Last edited:
Yes it has changed.