Peak Finder

hi all,
I'm currently working on a peak-valley (through) finder on EMG Data. But I can't automatically find peaks and valleys... Could somebody please have a look at my VI?
I Highlighted the area where I’m doing the peak valley analyses...
My Final goal is to take all the peaks as one point, and plot another line over those maximal datapoints to create a peak (or valley) line and calculate the slope of that function.
Does anybody perhaps has another way of doing this, or am I just totally out of track and doing the wrong thing?
Thanks for investing your time… J
Attachments:
Peak Valley another go.vi ‏1181 KB
post_mm40.txt ‏1800 KB

Hi MrHond (or MrDog),
Why didn't you connect the threshold and width (peak) inputs ?

Similar Messages

Peak detection in VC++?

Hello,
is there easy way to work with peak detector in VC++?
The class PeakDetector is available only in C sharp or VB. Of course, I could prepare a peak detector myself with second dt. But is there any way how use the Peak detector class in VC++?
Any advice or idea?
Thank You very much.
emta41

Hi,
Yes the VC++ version also has peak finding. Its called PeakDetector:
CNiPeakDetector m_PeakDetector;
m_PeakDetector.DetectPeaks(m_yData,CNiPeakDetector:eaks,m_fThreshold,m_nWidth);
However sometimes it doesnt find ALL the peaks, maybe due to the two params width and threshold.
Hope this helps a bit.
Miklos

How to measure the baseline of a noisy, pulsed signal

Hi
I am measuring the torque exerted by a large motor on a shaft using a load cell and lever arm. The shaft runs at approx 150 rpm. I have attached a drawing that shows the output I get. This is a test rig.
I have written some code that measures the maximum peak out of a group of approx 5 peaks and writes this to a shift register. This gives me an idea of the maximum torque "spike".
I also wish to measure the baseline torque (due to the bearings in the machine). Even when highly filtered (my noise filter is set to 49Hz) the signal exhibits this noise which is probably due to vibration in the system. The signal is zeroed when the motor is not running.
Does anyone have an ideas on how to measure the "baseline" torque? The large spike in torque prevents me from doing a running average. Can anyone think of a way of averaging just the noisy part of the signal to get an average value? I aim to to subtract the average baseline torque from the peak value to get an idea of the torque due to the event which causes
the spike.
Any help would be greatly appreciated.
Many thanks.
Attachments:
drawing of torque signal.gif ‏26 KB

Thanks for the reply. I understand what you are saying. However, I might have to modify my method for measuring the peaks if I choose to implement your idea. I have taken a screenshot of my "peak finder" code and attached it.
Bascially, the reset terminal is wired to a timer which outputs a pulse every few seconds. This resets the vi (a standard NI one I think) and sets the peak magnitude back to zero. This way, I am windowing the signal and measuring the maximum peak in every window. This is what I need to do.
So I could use a logical filter to feed data to the running average only if;
the amplitude of the signal is less than a certain threshold
and if the current value has similar low peaks either side of it
How would you construct the code to delay the evaluation so that the values in front and behind of the current data point can be analysed?
thanks again
Attachments:
peak_find_screenshot.jpg ‏45 KB

Slow running data for heart rate variability VI Please help my thesis is due in 3days!! :(

Hi there i'm having a great issue with a VI i have created to detect heart rate variabilty from ECG and PPG signals and seeing which one is most accurate with regards to HRV.
I initially recorded 25 minutes worth of data and saved this into a text file. which has 2 columns the 1st being PPG data the 2nd ECG. There are at least 1000000 rows of data
I am now running this data into a VI which should give me results such a RMSDD PNN50, poincare plots and histograms etc the trouble i'm having is the VI is initially starts of fast then becomes incredibly slow i left it running for over 10hours and it hardly read any data
and me not being such a big labview expert knows how to fix this. I would be most grateful if somone could help me find my error as my thesis is due in 3 days and i'm stuck on this last part. I have attached my VI below . please please help me!
Attachments:
CV Report1.vi ‏41 KB

First question: does the code produce the correct result, just slowly?
So you have a million data points and process them one point at a time in a FOR loop, updating several xy graphs and charts and building arrays. Most indicators will get overwritten with each iteration.
You are heavily filtering the data (twice!) and take the derivative (twice=second derivative). You are doing a pt-by-pt peak finding on 50 points with every new point added, meaning that you mostly analyze the same data (differing by only one point with each iteration. Seems inefficient. Wouldn't you detect most peaks 50x?
Since your data is filtered, maybe it would be sufficient to decimate the data before the loop processing. How correlated are adjacent points?
Have you analyzed which part of the code is slowest? What is the current execution time ("slow" is subjective) and what speed would you like to have instead?
Things like the histogram only needs to be done exactly once at the end. It seems pointless to do it over and over, a million times in a row even if the data did not change. Same problem for all the graphs.
You seems to take the difference between adjacent points in the generated array. Again you are repeating the same calculation over and over. Wouldn't it be much more effiicent to keep the array of pairwise differences in the shift register and use a scalar shift register to retain the immediately previous point. Subtract the new point from the previous and add it to the array instead.
LabVIEW Champion . Do more with less code and in less time .

Power Spectrum VI Performance

I am looking to accelerate the performance of the Power Spectrum and
Peak Finder VI's in LabVIEW. Right now to process 10000 points
the power spectrum takes 1.3 Miliseconds. While this number is
not huge, when doing 3000 of these analyses in a shot, it gets large,
quickly. I would like to know if there's anything to be done to
speed it up. For example, the nubmers i am sending it are fairly
small Single Precision Floats, would it work better with larger
numbers?...
Thanks for all your thoughts in advance.
Eugene

Hi
I read in some place, and time ago, that those vis (power spectrum,fft, psd, etc) are optimized for power of 2 number points. I am not sure if this keeps being true, but you can try to use, say, 8192 points, or zero pading to 16384 to see if speed increases.
Hope it helps.
Alipio
"Qod natura non dat, Salmantica non praestat"

Solve Linear Equations NaN return

I've implemented a modified Linear Levenberg Marquardt function for gauss fitting of spectral data, but have problems from time to time where I simply can't fit a specific data set.
I've traced the problem to the function "Solve Linear Equations" outputting NaN for one or more solution vectors (Not all) which then plays havok with my function.
I've located a data set which is initially OK, but which after solving the linear equations has NaN as the second element of the solution vector array. Changing the matrix type to something other than "generic" gets rid of the NaN for this particular data set, but gives problems with others which work with the "generic" matrix type.
Since my mathematics lectures in college are a few years away now, I'm asking someone out there if they can help me with my problem. I can supply a "corrupting" data set if this would help, or maybe someone knows this problem and has a work-around?
I've thought of leaving the matrix type to "generic" and checking the output to see if there's a NaN present, and if so, then re-calculate with another matrix type until there's no NaN. This is most likely not very sound mathematically, but it might reduce the number of rogue data sets.
Thanks in advance
Shane.
Using LV 6.1 and 8.2.1 on W2k (SP4) and WXP (SP2)

Hi Altenbach,
The data can have a significant amount of overlap. Typically three to four groups of peaks fully resolved, but each group can have up to 4-5 peaks in them, partially visible only as shoulders.
The start value "problem" is more or less solved (but as usual, not yet finished).
I've got an approximate first-guess peak find running which does a pretty good job of finding the peaks I need, and I then fit the height and FWHM first before doing a full optimisation (The Sub-VI approach allows much more flexible fitting models).
The fitting works well on almost all data sets. There are simply some spectra which (although visibly hardly different to others which work perfectly) do not fit, i.e. generate the NaN response from the "solve linear equations" function. It's a numeric problem (bug?), I'm sure, and not strictly a peak location or resolution problem. then again, I may be wrong.
I refer to the standard "LevMar" VI as being linear, because it assumes a linear relationship between the variable variation and the end mse used for optimisation. This is where the "Solve linear Equations" comes in. Since the relationship is almost certainly not linear (foe example when peaks overlap), I thought maybe the non-linear coefficient guess may yield better results. I've had a quick look through the non-linear LEV-MAR function, but don't understand it yet ot the extent I understand the linear one. It does indeed seem to take a slightly different approach (once you look past the whole parsing code of the "linear" function.
I'll need some time to get some understanding of the non-linear code.
Attached are some example spectra (One which works, and one which doesn't).
Thanks again
Shane
Using LV 6.1 and 8.2.1 on W2k (SP4) and WXP (SP2)
Attachments:
Spectra.zip ‏3 KB

PeakDetector/FreeAnalysisMem

I am trying to count the number of peaks on waveform, squared so as to
make the waves values postitive. I am only interested in the count
returned by PeakDetector,(PD). However I am unable to prevent it from
over allocating for the peak locations, amplitudes, and second
derivatives arrays. I've tried to do an iterative set up as exactly as
the example given with the Analysis package shows. One thing I've
found is that FreeAnalysisMem,(FAM) does not do much good. Also if the
pointers happen to have all elements with zero values, then it is
considered a NULL pointer and I get a Lib error from FAM. I've tried
realloc() on the ptrs. I realocated to zero, but something is still
allocated for this, and I eventually have no memory l
eft. I've tried
the free() method, but this apparently frees the memory so that it can
no longer be allocated. Why the CVI compiler does this I don't know,
but I've read "efficient C programming" and it claims that free() is
supposed to free the block so that it may be allocated again.
So, I am trying to use PD for the count only. If anyone can help me to
control the allocation taking place please help.
My algorithm is covering nine channels. at each channel I'd want to
scan, digitize, square, then find peaks at various levels. I then note
the count of peaks falling in a specific range(between two levels). I
have to call PD five times per channel. I've tried every combo of the
initialize and endOfData variable. Please help, or help me find a free
Peak finding algorithm somewhere out there that actually works in a
friendly sort of way.
THanks Jason
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.

I see what ou are saying about the CPU usage, but I think that you also have to look at the memory usage. LabVIEW 7.1.1 uses 43,524k of memory and LV 8.0.1 uses 24,568k, on my computer. I am on a computer with 3.4 GHz P$ with 2 gb of ram.
Although LV 8 uses more cpu time, it uses hlaf the memory so there is a trade off there.
I am not sure how you can make the CPU time decrease or the memory increase for LV 8. The programming is the same in both your vis. I will do some more tsting when I get a chance and reply back
Kenny
Kenny

Datenrate in bestimmten Bereichen ändern

Hallo,
ich möchte in bereits vorhandenen Datenaufzeichnungen die Datenrate für bestimmte Bereiche ändern um die Datengröße zu reduzieren aber die Kurve dabei nicht verkürzen.
Beispiel:
   es werden Daten über eine Zeit von 60 Sekunden erfasst. Zwischen einzelnen Peaks sind längere Pausen wo die Spannung zwischen 0 und x Volt liegt. Diese Bereiche möchte ich gerne kürzen.
Hatte an ein VBS Script gedacht aber da stoße ich immer wieder auf Probleme, da die Länge des Kanals bis zu 49247232 Werte umfassen kann und mehr.

smooth
create_groups
if Data.Root.ChannelGroups(1).Channels.Exists("PXI1Slot2\ai0") Then
    find_peak_hv1
    kurve_teilen_hv1
end if
if Data.Root.ChannelGroups(1).Channels.Exists("PXI1Slot2\ai2") Then
    find_peak_hv2
    kurve_teilen_hv2
end if
'Kurven glätten
sub smooth
if Data.Root.ChannelGroups(1).Channels.Exists("PXI1Slot2\ai0") Then
    Call ChnSmooth("[1]/PXI1Slot2\ai0","[1]/u_hv1",2,"maxNumber") '... Y,E,SmoothWidth,SmoothType
End if
if Data.Root.ChannelGroups(1).Channels.Exists("PXI1Slot2\ai1") Then
    Call ChnSmooth("[1]/PXI1Slot2\ai1","[1]/i_hv1",2,"maxNumber") '... Y,E,SmoothWidth,SmoothType
End if
if Data.Root.ChannelGroups(1).Channels.Exists("PXI1Slot2\ai2") Then
    Call ChnSmooth("[1]/PXI1Slot2\ai2","[1]/u_hv2",2,"maxNumber") '... Y,E,SmoothWidth,SmoothType
End if
if Data.Root.ChannelGroups(1).Channels.Exists("PXI1Slot2\ai3") Then
    Call ChnSmooth("[1]/PXI1Slot2\ai3","[1]/i_hv2",2,"maxNumber") '... Y,E,SmoothWidth,SmoothType
End if
end sub
'Untergruppen erstellen
sub create_groups
Call Data.Root.ChannelGroups.Add("U_HV1")
Call Data.Root.ChannelGroups.Add("I_HV1")
Call Data.Root.ChannelGroups.Add("U_HV2")
Call Data.Root.ChannelGroups.Add("I_HV2")
end sub
'Kurven Start- und Stop-Werte finden für HV1
sub find_peak_hv1
dim oChn, i, low_hv1, high_hv1
dim x
Dim peak, low, st
x = 1
st = 1
Call Data.Root.ChannelGroups(1).Channels.Add("high_hv1",DataTypeFloat64)
Call Data.Root.ChannelGroups(1).Channels.Add("low_hv1",DataTypeFloat64)
set oChn = Data.Root.ChannelGroups(1).Channels("u_hv1")
set high_hv1 = Data.Root.ChannelGroups(1).Channels("high_hv1")
set low_hv1 = Data.Root.ChannelGroups(1).Channels("low_hv1")
While low <> 0 or peak <> 0 or x <= 100
    low = Find("Ch(""u_hv1"")>=3",st)
    if low >= 0 then
      peak = Find("Ch(""u_hv1"")<=3",low)
      if low >=0 and peak >= low and peak-low >=10 then
        low_hv1(x) = low
        high_hv1(x) = peak
        st = peak
        x = x + 1
      else st = peak
      end if
    end if
Wend
end sub
'Kurven Start- und Stop-Werte finden für HV2
sub find_peak_hv2
dim oChn, i, low_hv2, high_hv2
dim x
Dim peak, low, st
x = 1
st = 1
Call Data.Root.ChannelGroups(1).Channels.Add("high_hv2",DataTypeFloat64)
Call Data.Root.ChannelGroups(1).Channels.Add("low_hv2",DataTypeFloat64)
set oChn = Data.Root.ChannelGroups(1).Channels("u_hv2")
set high_hv2 = Data.Root.ChannelGroups(1).Channels("high_hv2")
set low_hv2 = Data.Root.ChannelGroups(1).Channels("low_hv2")
While low <> 0 or peak <> 0 or x <= 100
    low = Find("Ch(""u_hv2"")>=3",st)
    if low >= 0 then
      peak = Find("Ch(""u_hv2"")<=3",low)
      if low >=0 and peak >= low and peak-low >=10 then
        low_hv2(x) = low
        high_hv2(x) = peak
        st = peak
        x = x + 1
      else st = peak
      end if
    end if
Wend
end sub
'Kurven teilen HV1
sub kurve_teilen_hv1
dim u_hv1_ch, i_hv1_ch
dim x
Dim low_hv1, high_hv1, st
x = 1
set high_hv1 = Data.Root.ChannelGroups(1).Channels("high_hv1")
set low_hv1 = Data.Root.ChannelGroups(1).Channels("low_hv1")
for x = 1 to Data.Root.ChannelGroups(1).Channels("high_hv1").Properties("length").Value
    u_hv1_ch = "uhv1_" & x
    i_hv1_ch = "ihv1_" & x
    Call Data.Root.ChannelGroups("U_HV1").Channels.Add(u_hv1_ch,DataTypeFloat64) 'hier entstehen Numerische Channel, diese sollen aber den Zeitstempel übernehmen
    Call Data.Root.ChannelGroups("I_HV1").Channels.Add(i_hv1_ch,DataTypeFloat64)
    Call DataBlCopy(Data.Root.ChannelGroups(1).Channels("u_hv1"),low_hv1(x),high_hv1(x)-low_hv1(x),Data.Root.ChannelGroups("U_HV1").Channels(u_hv1_ch),1)
    Call DataBlCopy(Data.Root.ChannelGroups(1).Channels("i_hv1"),low_hv1(x),high_hv1(x)-low_hv1(x),Data.Root.ChannelGroups("I_HV1").Channels(i_hv1_ch),1)
next
end sub
'Kurven teilen HV2
sub kurve_teilen_hv2
dim u_hv1_ch, i_hv1_ch
dim x
Dim low_hv2, high_hv2
x = 1
set high_hv2 = Data.Root.ChannelGroups(1).Channels("high_hv2")
set low_hv2 = Data.Root.ChannelGroups(1).Channels("low_hv2")
for x = 1 to Data.Root.ChannelGroups(1).Channels("high_hv2").Properties("length").Value
    u_hv1_ch = "uhv2_" & x
    i_hv1_ch = "ihv2_" & x
    Call Data.Root.ChannelGroups("U_HV2").Channels.Add(u_hv1_ch,DataTypeFloat64)
    Call Data.Root.ChannelGroups("I_HV2").Channels.Add(i_hv1_ch,DataTypeFloat64)
    Call DataBlCopy(Data.Root.ChannelGroups(1).Channels("u_hv2"),low_hv1(x),high_hv1(x)-low_hv1(x),Data.Root.ChannelGroups("U_HV2").Channels(u_hv2_ch),1)
    Call DataBlCopy(Data.Root.ChannelGroups(1).Channels("i_hv2"),low_hv1(x),high_hv1(x)-low_hv1(x),Data.Root.ChannelGroups("I_HV2").Channels(i_hv2_ch),1)
next
end sub
Hier das Script. Beim erstellen der Channel, in die die Kurven übertragen werden, entstehen Numerische Kanäle. Diese haben allerdings keinen Zeitstempel sondern verwenden als X-Skalierung die Zeilennummern. Wie erstellt man auch dort eine Zeit-Achse?
Können Sie mir anhand diesem Beispiel sagen wie ich nach dem kopieren Werte verwerfen kann und damit eine dynamische Datenrate erzeugen kann?
Vielen Dank
Sven Hermsmeier

Stepped signal analysis

how to count the number of steps(red signal on graph) on the diagram below?
number of steps should be counted in the rising and falling signal and make a comparison between them.
how this can be done?
Attachments:
diag.jpg ‏55 KB

There are probably lots of ways. Your diagram has a hint for one of them. Here are a couple of options:
Use a Savitzky-Golay filter to take the derivative of the waveform. You will end up with a bunch of peaks and valleys. Find the peaks with a peak finder and count them. You can figure out up/down by looking at the area between the peaks and get an average level or by looking at the direction of the peaks (up is up, down is down). You data is fairly noisy, so you will have to experiment with how wide you want the S-G filter to be and how much averaging you want to do between the peaks.
Use a single cycle square wave as a convolution filter and run it. The end result will be similar to the Savitzky-Golay filter and can be analyzed the same way. This is probably easier and faster on some targets, such as an FPGA.
Good luck. Let us know if you need more help.
This account is no longer active. Contact ShadesOfGray for current posts and information.

How can the coordinates of an intersection point of a cross

there is a cross on a chart and i need the coordinates of the intersection point as a reference for further measurements

The simplest way is pattern matching. Train the routine with the cross, then search for it. It should always be within about 1 pixel of the intersection. You can use sub-pixel accuracy to improve the results sometimes.
To improve the pattern matching results, you can use edge or peak finding routines to find the horizontal and vertical lines that make up the cross. You can determine the center from these lines.
Bruce
Bruce Ammons
Ammons Engineering

Where can i find a Labview driver for the "8990a PEAK POWER ANALYZER" it's not on NI's site.

where can i find a Labview driver for the "8990a PEAK POWER ANALYZER" it's not on NI's site.
if there is no driver maybe you can tell me the closest instrument that comes to this one from the list at,
http://search.ni.com/query.html?lk=1&col=alldocs&nh=500&rf=3&ql=&pw=595&qp=%2BContentType%3AInstrumentDriver+%2BIDNetManufacturer%3A%22Hewlett-Packard%22&qt=peak+power+analyzer&layout=IDNet
so i could modify and use the driver

Hello,
It looks like Agilent/HP makes the 8990A, but I was not able to find a driver for it that we or Agilent has written for this device.
Since I am not familiar with the 8990A, I cannot recommend which drivers might be similar, but you are pointed in the right direction. Agilent themselves might know more about which models use similar command sets so that you can modify a driver easily.
Hope this helps.
Scott B.
Applications Engineer
National Instruments

How to find max audio peak of combined audio tracks with effects applied?

Hello all,
I am working in PPro CS5.5.
Often when I am working on a timeline I have two audio tracks of the same songs, one from the camera and one from a separate sound recorder, synced on the timeline so they combine into one signal. Applied to these tracks I might have an E.Q. effect or highpass/lowpass filter. Now I know that the max peak of each track or individual clips is easy to find using the 'Audio Gain' menu item, what I am hoping for is the equivalent of that menu item but related to the final output after the two tracks signal has been added together and effects have been all been applied? It is a very laborious task to have to listen through the whole timeline to discover if the output is going over 0db, if there was a way I could locate on the timeline where the max output is, that would definitely solve the problem.
Finding the way to do this would save literally hours per project! It's the one thing that makes me inwardly groan when I have to effect the output volume, which I have to do most of the time.
Hopefully,
Peter.

I have found audio to be more difficult to learn than video. Audio has so much more subtlety, where perceptions don't always equate to what seems obvious. With multiple tracks adding together things get very complex very fast and peak amplitude has very little relationship to volume. When you compress a waveform the volume can be increased! Who'd have thunk it? Then you have S/N ratios and harmonics, soft clipping and hard clipping... Which brings me to the option on the Dynamics Effect called Soft Clip that is a smoother kind of limiter.
It's amazing where a simple one line suggestion can lead you; having worked a bit more with the dynamics 'effect' I am deciding that I like it a lot. The automation is not absolute but it broadens the tolerances I can work within. Very liberating!

How to calculate the slope of a signal to find its peak

I have acquired a ECG signal through the DAQ. Actually I have to calculate the slope of the ECG signal in order to find the R-peak. By finding the slope of R- wave which will be greater than the slope of T-wave, I could differentiate between R-wave and T-wave as it looks similar for some patients. Can any one of you guide me in finding the slope of the incoming signal and then detect the R-wave.
Thank you.

sorry, I do not have version 7.1, so I am attaching it in LV8.0 and a picture of the block diagram. Maybe someone can convert it for you, or that you can see enough to recreate the code from my pics... not too much going on. Hope this works for you.
Message Edited by Stradis on 10-11-2007 04:21 AM
Paul <--Always Learning!!!
sense and simplicity.
Browse my sample VIs?
Attachments:
trigger to sample.png ‏10 KB
trigger to sampleWithSlope.vi ‏288 KB

Is there a way to find peak audio level?

I'm using FCE 4 and I am trying to adjust the audio level on a number of tracks. In most of the I can do it without problem using Normalization Gain, but one track refuses to adjust. From what I have read it could be because there is a peak (audience clapping) at the end of the track. If I split the track into a different clip item, should that work? or is there a way of finding out where a peak is without listening to the entire track? I saw something like that but it may have only applied to FCP.

I don't believe that there's a way to find the peak (or peaks) within FCE. But if the track isn't normalizing then there should be at least one peak somewhere.
However, I have experienced audio anomalies when playing back and when exporting content with four to six audio tracks: at certain moments, the volume rises abruptly, even though such a volume increase is not present in any of the audio tracks. Perhaps FCE is incorrectly "reading" a volume increase at a certain moment, thus interpreting it as a peak, and therefore "deciding" that normalization is not possible.
By the way, my solution for fixing the audio anomaly that I mentioned above was to process my audio in an external editor, as I simply could not get FCE to handle it properly, even after trashing my preferences files, trashing all of my render files, creating a new Sequence and copying my clips to it, etc.

Finding peaks in an ECG-signal

How can I find the T-wave and P-wave in an ECG-signal? Can I do a bach/forward search from the QRS-peak?
Thank you!

One method would to find the peaks of T wave, QRS and P wave.
Then probably, you can index out the relevent section of the waveform that will comprise your T and P wave
Look at an example vi I have attached
Attachments:
DETECT_PEAKS.vi ‏27 KB

Peak Finder

Similar Messages

Maybe you are looking for