From: James Brown (Jim_at_Seti.Net)
Date: Sun Feb 18 2007 - 10:25:24 PST
I started a conversation on the Borland Delphi graphics group when I asked for advice in finding signals in the JPGs that I am creating and thought it might be interesting to the SETI community as well.
I appreciate all the help that has been offered to me in this task but I feel that I failed to make my needs clear. Let me try again.
I am searching for a beacon from outside the earth and success will be when I can identify such a signal and show, with a certain amount of probability that the signal is not from a local source. The end result will be the frequency of the signal and pointing angles to the source - just those two things.
During the search I record the output of my receiver audio in WAV format (raw PCM) because it's lossless. I also attach to the WAV files all the engineering information in the form of an XML section. This way if a signal is thought to be present I will know where I was looking when it happened. This part is critical because the search space is huge. For my small station its about 12 million combinations of frequency and pointing angles.
An important part of the detection process is conversion of the raw PCM data to a form that lends itself to analysis. I chose to do this by first performing an FFT transform and then a waterfall display of the data. This allows my own human ability as a superb pattern recognition engine to quickly identify any lines in the waterfall. I then save the waterfall bitmap image as a JPG file and also tag it with the XML engineering data. The idea is that if a line shows up while I'm not staring at the waterfall display it will still be saved in the JPG along with tags that relate back to the raw PCM data.
The problem is that I can't simply stare at the waterfall display for endless hours and must rely on reviewing the JPGs for lines. The secondary problem is that I can't even stare at the thousands of JPGs without hallucinating images of all sorts in the noise.
What I need is a detector that will ring an alarm bell when a line is present in the bitmap image of the waterfall. This could be a technique applied directly to the waterfall in real time or after the fact during processing the JPG files. If the JPG format will cause possible lines to be lost, because it attenuates some of the components of the image and PNG format would be better at the task I can switch to that format.
The best of all solutions would be a real time detector that worked on the Bitmap image before it is saved in either JPG or PNG.
Argus Station: DM12jw
This archive was generated by hypermail 2.1.6 : Sun Feb 18 2007 - 10:31:37 PST