fft based filtering with FIR windows!

gistfile1.py

import numpy as np
import scipy.signal
import matplotlib.pyplot as plt

n=256
data = np.random.random(n)
data_fft = np.fft.fft(data)

# see also : http://mpastell.com/2010/01/18/fir-with-scipy/

# low pass
a = scipy.signal.firwin(256, cutoff=0.5, window='hanning')
# high pass
b = scipy.signal.firwin(256, cutoff=0.3, window='hanning'); 
b[n/2] = b[n/2] + 1
# combine
d = - (a+b); 
d[n/2] = d[n/2] + 1
# fft, filter, ifft
H = np.fft.fft(b)
data_filt = np.fft.ifft(data_fft*H)  # note that you could rotate this around for a 2d window
plt.plot(data)
plt.plot(data_filt)

# you can also filter really easily with scipy.signal.filtfilt()
data_filt = scipy.signal.filtfilt(b=d, a=[1], x=data)