Hello, I am trying to detect mutiple peaks in a FFT. 90% of the time the primary peak (Highest magnitude) is the only return i am interested in. But sometimes i have multi peaks and i need a way to detect these peaks. It gets tricky especially with the noise in the fft resembles mini-peaks. I get multi peaks when i have reflections in a signal, and i want to determine these reflection points when they happen. Thanks in advance! GC
Detecting Multi Peaks in a FFT
Started by ●July 16, 2007
Reply by ●July 16, 20072007-07-16
On Jul 16, 5:57 pm, "gchamsine" <gus_...@hotmail.com> wrote:> Hello, I am trying to detect mutiple peaks in a FFT. 90% of the time the > primary peak (Highest magnitude) is the only return i am interested in. > But sometimes i have multi peaks and i need a way to detect these peaks. > It gets tricky especially with the noise in the fft resembles mini-peaks. > I get multi peaks when i have reflections in a signal, and i want to > determine these reflection points when they happen. > Thanks in advance! GChmm... the any jump in value greater than from TBD above the 'nominal value' can be qualified as a peak? Btw, how does reflections result in generation of additional frequency components. If you have any pointers, it might be helpful. Krishna http://dsplog.blogspot.com
Reply by ●July 16, 20072007-07-16
>On Jul 16, 5:57 pm, "gchamsine" <gus_...@hotmail.com> wrote: >> Hello, I am trying to detect mutiple peaks in a FFT. 90% of the timethe>> primary peak (Highest magnitude) is the only return i am interestedin.>> But sometimes i have multi peaks and i need a way to detect thesepeaks.>> It gets tricky especially with the noise in the fft resemblesmini-peaks.>> I get multi peaks when i have reflections in a signal, and i want to >> determine these reflection points when they happen. >> Thanks in advance! GC > >hmm... the any jump in value greater than from TBD above the 'nominal >value' can be qualified as a peak? >Btw, how does reflections result in generation of additional frequency >components. If you have >any pointers, it might be helpful. > >Krishna >http://dsplog.blogspot.com > > >I am using a laser based radar to determine range, and when the laser bounces off multiple surfaces within the depth of focus of the laser, i see peaks in the fft for each surface the laser bounced off. Thats where i need to determine a method to define "what is a peak" that is robust enough.If i just look at the highest magnitude peak, then i may return the "reflected peak" which i don't want. And returning the max peak is easy, but when i want to determine how many peaks exist it gets tricky, i know i need to determine a threshold and say everything above that line is considered peaks...but that doesnt work for all cases, i need a more robust way. Thanks in advance for ur sugestions!