Reg : FFT Filter

Hello, I am just a beginner in Matlab, and just started to explore the complete usage of MatLab. I am given a task on filtering a noisy data. Say for Ex: I am given a inputdata.mat file, i loaded and plotted the data variable inside the mat file. got a curve with lots of noise. I want to use the FFT filter to filter the noisy data. As am a beginner, i would really appreciate if anyone can help me out in this. I know i have to use the fft command for the same, but i couldnt move any further. Need your help Thanks Venkat

Answer 1

[William]

What version are you using? If you are using the student version you can use the "Filter" command within the control systems toolbox to visually create a filter.

Next you need to figure out how your data is input. If your .mat file is just points of a random noise function stacked on top of a sine wave you can use the filter function

help filter

If you don't have the control systems toolbox you can find the transfer function and work the problem with a number of for loops in a c code type manner.

If you can at all use the "filter" command. It does all the FFT work for you.

Answer 2

[Venkat Varadarajan]

Hello William, Thanks for help. Am using R2010a. Actually my data is just a set of values from .mat file. i just loaded my mat file and when i plot the data inside that, i get a wave with lots of noise. My aim here is to reduce the noise thats it.

Can you help me with just an algorithm for the same ? I prefer doing in mscript than simulink.

Any filter is fine for me. Like Lowpass filter, FFT or Kalman. Any advice on this please.

Answer 3

[Wayne King]

If you have the Signal Processing Toolbox, you can:

1.) specify your filter using fdesign.lowpass (for example)

2.) design your filter using design()

3.) apply your filter to data using filter() or filtfilt()

In this example I design a lowpass FIR equiripple filter for data sampled at 10 kHz. I'll pass everything below 1 kHz, with the stopband starting at 1100 Hz. I specify 60 dB of attenuation in the stopband and 0.5 dB of passband ripple.

I'll create some noisy signal first to filter.

     t = 0:1/1e4:1;
     x = cos(2*pi*250*t)+sin(2*pi*500*t)+randn(size(t));
     %Now to specify the filter
     Fs = 1e4;
     d = fdesign.lowpass('Fp,Fst,Ap,Ast',1000,1100,0.5,60,Fs);
     % use specifications to design the filter
     Hd = design(d,'equiripple');
     % apply the filter to the noisy signal
     y = filter(Hd,x);

You can view your filter response with:

fvtool(Hd)

Hope that helps,

Wayne