Extracting frequencies from multidimensional FFT

问题:

I've written a python code to take a 2D signal and FFT it, and now I want to extract the frequencies associated with the FFT. The np.fft.fftfreq fails, giving me the error

File "/usr/lib64/python2.7/site-packages/numpy/fft/helper.py", line 153, in fftfreq
    assert isinstance(n,types.IntType) or isinstance(n, integer)
AssertionError

My code is :

import numpy as np
import scipy as sp
import pylab
import sys
import math

filename = sys.argv[1]  # Get name of file to open 

ifp = open(filename, "r")
ifp.seek(0)

nrows = 0
ncols = 0

nrows = sum(1 for line in ifp) # Sum over all the lines in the file ptr

ifp.seek(0) # Set the fptr back to beginning of file
for line in ifp:
   ncols = len(line.split()) #Split and count number of words in a line
   if ncols > 0:
      break

OrigData = np.zeros([nrows, ncols], dtype=np.float32) #Allocate numpy array
FFTData = np.zeros([nrows, ncols], dtype=complex)
IFFTData = np.zeros([nrows, ncols], dtype=complex)
FreqComp = np.zeros([nrows, ncols], dtype=np.float32)

ii = 0
jj = 0
ifp.seek(0)
for line in ifp:
   linedata = line.split()
   jj = 0
   for el in linedata:
      OrigData[ii,jj] = float(el)
      jj+=1
   ii+=1
ifp.close()

FFTData = np.fft.fft2(OrigData)
FreqComp = np.fft.fftfreq(FFTData, d=2)

#--- Continue with more code ---#

I know that everything else works except the np.fft.fftfreq line, because I added that in last. How does one extract 2 dimensional frequency components?

回答1:

You are passing in an invalid parameter: np.fft.fftfreq takes the size of the signal data as first parameter (an integer) and the timestep as the second parameter. You are passing in an array as the first parameter.

You need to perform an np.fft.fft on the signal first though.

Hate to point out the obvious, but read np.fft.fftfreq... the example code is very pretty clear.

---

Having performed a 2D FFT, you can obtain the sample frequencies along each dimension as follows:

FreqCompRows = np.fft.fftfreq(FFTData.shape[0],d=2)
FreqCompCols = np.fft.fftfreq(FFTData.shape[1],d=2)