Add noise to artificial data???

[Hyun-Seob, Lee]

Dear all.

I am studying effect of filter.
Artificially, I made raw data as function by time used excel.
I must input noise values to raw data for reality but i don't know method about noise process.
I would really appreciate if anyone introduce to me about mathematical or statistical methods which is add noise values to the raw data.

Thanks,
Lee, HyunSeob

[sivanantham]

Re: Add noise to artificial data???

what type of noise data you want to add to your raw data & what is your data?. ex. power line noise in respiration signal

[Hyun-Seob, Lee]

Re: Add noise to artificial data???

First of all, thanks sivanantham for reply to my post.

I have data about gait(artificial data).
Generally when we got gait data, the data is included noise. therefore we are filtering the data in order to remove noise.
But my data have no noise because made by artificially.
I want to add randomized noise in 5% of data approximately.

[Manvendra Singh]

Re: Add noise to artificial data???

There are functions in excel for generation of random nos. This link should help you using Excel to add noise http://www.pcreview.co.uk/forums/add...-t2948315.html

[sivanantham]

Re: Add noise to artificial data???

If you are using MATLAB try this to generate & add noise to your gait data.

nvar = .50; % Noise variance
noise = 2*randn(200,1)*nvar; % Generate Random noise or
noise = wgn(200,1,0);

NoiseData = OrignalData+noise;

Use either wgn() or randn() in matlab to generate noise .The array length of both Noise and original data should be same.

If you don't have MATLAB try RAND() function in excel to generate the noise data.

[Hyun-Seob, Lee]

Re: Add noise to artificial data???

Thanks for your response..

[Hyun-Seob, Lee]

Re: Add noise to artificial data???

Thank you for your response....

[Ted Andresen]

Re: Add noise to artificial data???

Hi HyunSeob

You might do some thinking as to whether you need to corrupt your data with additive or multiplicative noise. The first would be GaitData(t) + NoiseAmp * Gaussian, while the second might be GaitData(t)*0.05*Gaussian. I don’t know which one the filter is supposed to remove.

In Excel just adding three, Rand(), functions together will generate a pretty acceptable example of Gaussian noise. “Rand()+Rand()+Rand()” has a mean of 1.5 and a sigma of 0.5.

You can use the sum of three Rand()’s in a multiplicative corrupting function or noise as

=1+(RAND()+RAND()+RAND()-1.5)/10

This one has a mean of ~1 and an SD of ~0.05 or 5%.

You can see how it behaves if you paste the above expression in cells A1 to A1000, and paste =Average(A1:A1000) in cell A1001 and =STDEV(A1:A1000) in cell A1002 Then you can repeatedly press F9 and watch Excel recalculate the mean and sigma. That’s fun.

If you want an additive corrupting function try:

=NoiseAmp *((RAND()+RAND()+RAND()-1.5))/20

Play with the expressions and see what type of corrupting function meets your needs.

Ted Andresen
St. Petersburg, FL
USA

[William Rose]

Re: Add noise to artificial data???

Hyun-Seob,
Ted Anderson correctly suggests using rand()+rand()+rand() in Excel to generate a random number whose distribution is closer to normal than what you'd get if you just rand() alone. But if you want "normal" noise (and it's a good idea), just create it exactly in Excel, as follows:
=NORM.INV(RAND(),0,1)
The above Excel formula has the added advantage that the user directly specifies the desired mean and standard deviation (0 and 1 respectively, in my example).
Bill R.