Search code examples
algorithmsignalsidentification

Algorithm for signal identification

I am getting continuously 100 k data points per second with a signal noise ratio (SNR) about 0.1~1%. Is there any FAST/REALTIME algorithm to identify these signals? The simple histogram helps for higher SNR, but not for 0.1%

The plot gives an impression how my data looks like, but with higher SNR.

Thanks for any inputs!!

enter image description here

Solution

  • There is a "not too bad" solution now, see result from the attached plot final result

    (1) collect coming points (data or noise) and fill up #1 array; (2) for the next coming point, check whether there are enough friends (eg. 3; friend means the difference between them smaller a threshold) in #1 array; if yes mark this point as an identified; collect all identified points and fill up in #2 array; (3) repeat 2nd step but compare with items in #2 array, and then fill up all identified points into #3 array;

    I plot all points in #3 array, and it seems that the identification quiet good especially with the help of histogram -- only show the peak.

    Plot: left -- original points; middle -- identified points; right -- the peak of histogram.