Peak detection algorithm

We decided that a hueristic approach to an adaptive threshold could be using a pdf of a wider band than the one we are sensing. The idea is that we assume the noise energy is prominantly feature on the lowest part of the energy range. Therefore we needed to find the peak of the noise - mean of the main noise distribution.

After considering use of the TSpectrum class from ROOT I found a simple Matlab function that seemed to do what I wanted.

http://billauer.co.il/peakdet.html

Actually I used a Python version of the script I found posted at StackOverflow.

http://stackoverflow.com/questions/1713335/peak-finding-algorithm-for-python-scipy

Which I then converted to C++ in gnuradio. So far after playing with some of the input parameters it seems to work OK. Since we assume that the noise energy is going to be prominently featured at the low-end of the scale, this algorithm appears accurate enough to identify that noise peak.