FFT of EMG Signal

I am trying to obtain the FFT of an EMG Signal, I have already done some other type of processing, including; removing DC Offset, Rectification and Signal Envelope, however I am stuck in the FFT part ? Any help would be appreciated ..

Attached the Code and Photo of current situation can be found.

[dpb]

"I am stuck in the FFT part"

For what working definition of "stuck", pray tell?

[Karl Rueth]

For some reason, the final result, isn't the required signal.

[dpb]

What "final result", there's no FFT uncommented in the sample code; the result will be what it shall be; you can't really "require" anything.

You don't even say what your desired result is that you're looking for, specifically, what are we, mind-readers?

Look at the example at

doc fft

for example of doing PSD via FFT for pointers.

Answer 1

[Star Strider]

See the documentation on fft (link).

Answer 2

[Luc Dransfeld]

If you follow the FFT documentation from Matlab, you should be able to get there. Maybe though, you should set your plot window differently, as motion artifacts can contain a power that is proportionately way bigger that the rest of the EMG-frequencies resulting that you will only visibly see the motion artifacts. You can always use matlab function: periodogram, which basically gets you the logarithm from the power, expressed in dB.

Answer 3

[Trishna Sah]

This is using your variables in the fft doc by Matlab

%fft plot for emg4
Fs=1000;
L=length(emg4);
t=L/Fs;
x=emg4;
Y = fft(x);

P2 = abs(Y/L);
P1 = P2(1:L/2+1);
P1(2:end-1) = 2*P1(2:end-1);

f = Fs*(0:(L/2))/L;
plot(f,P1) 
title('Frequency Spectrum Plot')
xlabel('f (Hz)')
ylabel('|P1(f)|')

f = Fs*(0:(L/2))/L;
plot(f,P1) 
title('Single-Sided Amplitude Spectrum of X(t)')
xlabel('f (Hz)')
ylabel('|P1(f)|')

[Dominic Rasp]

Hi Trishna,

could you describe what you do with P1 and P2 and why you do this?

Also what unit is the Y-Axis? It should be the number of signals at the particular frequency, right? How come, I get values of 0.12 at the Y-Axis?

Cheers

[dpb]

P2 is two-sided (-Fmax - Fmax) PSD.

P1 is one-sided (0:Fmax) PSD

Since the total energy is distributed over the entire P2, P1 is then normalized to conserve energy by doubling all components except DC and Fmax which are unique and full-valued in P2.

This normalization produces peak values that match the time series magnitude at the given frequency -- it will return 1.0 for a base sin() input of a frequency that matches the bin spacing of the FFT precisely, otherwise one will have to integrate (sum) over the bins containing energy around the input frequency to get the precise total.

This is illustrated by the example in doc fft from which the above code was extracted.

The comment about "number of signals" appears to indicate some confusion about what a PSD represents. As the example shows, it returns the magnitude (or energy by alternate scaling) of the input signal at the given frequencies. It isn't clear what you expected to have gotten...

[Dominic Rasp]

Thanks for the reply - got you so far. So when fft-ing an EMG signal, the Y-axis in your plot would be Volts (or the respective original signal unit), right? If I wanted to convert that into dB, I would have to 20*log10 the absolut fft signal - is that correct?

[dpb]

Ayup.

[ANADI BISWAS]

Units of x axis and y axis after abs(fft)?