# Peak detection for signals that contain high frequencies

I have an question about peak detection. My purpose is creating a plot with an impulse response and the decay line at / above it.
What I see till now is that there are two options for finding peaks:

• the simplest version is to just compare point with the points beside it $f(n-1) , f(n) , f(n+1)$
• The other assumes that there are at least 3 or 4 points per half wave (DSP related peak detection)

But my simple self recorded sweeps and transfer function generate just a impulse response with high frequencies. So the peak with option 1 is a peak of the highest sample value but isn't necessarily the highest audio peak.
Also the situation with 3 or 4 samples don't exist.

I'm not looking to a complete answer or work around(however examples are fine). Some links or names from papers would be nice.

• I am not sure I understand your question. "Creating a plot", "decay line", "finding peaks", "high frequencies", "sweeps" are all terms I am familiar with, but I cannot put the puzzle together. Perhaps a hand-drawing of what you really wish? – Laurent Duval Oct 18 '16 at 21:18
• I add a picture where in I try to explain what i want – Jan-Bert Oct 19 '16 at 6:40
• What is the background of that? Do you want to match exponential decay to your IR? – jojek Oct 19 '16 at 8:38
• Yes and maybe start point and end point for RT60. but first i have to find out if that's aloud... – Jan-Bert Oct 19 '16 at 10:27
• OK, so you want to take whatever IR you have and calculate RT from it? And for that you want to match the decay? If so, then I have an easy solution for you. – jojek Oct 20 '16 at 6:45