Anyone know how to calculate median frequency?

[Timmy]

Hi,

I am trying to rig up a makeshift EMG in my research group using LabVIEW 8.0 and a DAQ card, and I was wondering if there was a function, or a set of functions that I could use to determine the median frequency of the EMG. Looking at signals theory, the median frequency is defined as the frequency at which the integral of the power spectrum is equal to half of its maximum value. Right now, the way I'm trying to program that in LabVIEW is by using the Power Spectrum Express VI wired to the Integration express VI. From there, I convert the integral to a floating point value and divide it by two. What I don't know is how to back-correlate the integral value to a specified frequency. If someone could clue me in on how to do this, or suggest another approach for finding the median frequency, it would be much appreciated.

Thanks,

Timmy

[mross]

QUOTE (Timmy @ Apr 9 2008, 02:50 PM)

Hi,

I am trying to rig up a makeshift EMG in my research group using LabVIEW 8.0 and a DAQ card, and I was wondering if there was a function, or a set of functions that I could use to determine the median frequency of the EMG. Looking at signals theory, the median frequency is defined as the frequency at which the integral of the power spectrum is equal to half of its maximum value. Right now, the way I'm trying to program that in LabVIEW is by using the Power Spectrum Express VI wired to the Integration express VI. From there, I convert the integral to a floating point value and divide it by two. What I don't know is how to back-correlate the integral value to a specified frequency. If someone could clue me in on how to do this, or suggest another approach for finding the median frequency, it would be much appreciated.

Thanks,

Timmy

Take the derivative?

[Anders Björk]

Sorting the amplitudes and taking the value at 50% or if no value exist at 50% do an interpolation between value below and above 50%. Now you would be able to the frequency indexes and interpolate.

[shoneill]

QUOTE (Anders Björk @ Apr 10 2008, 09:12 AM)

Sorting the amplitudes and taking the value at 50% or if no value exist at 50% do an interpolation between value below and above 50%. Now you would be able to the frequency indexes and interpolate.

Integrate over the entire power spectrum. (You get an ever-increasing curve as a result).

Then look for the index which corresponds to (CurveMAX/2) and if neccessary interpolate between two individual data points.

Find the frequency associated with that index (or interpolated index) and there you have it!

Shane.