Ruby FFT of .wav Test.rb

#!/user/bin/env ruby

require 'rubygems'
require 'wavefile'
require 'fftw3'
require 'fastercsv'

w = WaveFile.open('sample.wav')

samples = w.sample_data[0, [w.sample_rate * 10, w.sample_data.size].min]
duration = samples.size / w.sample_rate

na = NArray.float(2, samples.size)
samples.each_with_index do |v, i|
  na[0, i - 1] = i.to_f / w.sample_rate.to_f
  na[1, i - 1] = v
end

fa = FFTW3.fft(na)
fa = fa.real.abs # you probably need absolute values for magnitudes of a frequency to make sense

FasterCSV.open('output.csv', 'w') do |csv|
  0.upto((fa.total / fa.dim) - 1) do |i|
    csv << fa[true, i].to_a
  end
end

Ran into a weird bug using this code. FYI, if you get this error:

fouriertest.rb:16:in []=': end-index=2 is out of dst.shape[0]=2 (IndexError) from fouriertest.rb:16 from /Users/dbrady/.rvm/rubies/ruby-1.8.7-p249/lib/ruby/site_ruby/1.8/rubygems/custom_require.rb:36:ineach_with_index'
from fouriertest.rb:14:in each' from fouriertest.rb:14:ineach_with_index'
from fouriertest.rb:14

It's because the wavefile you're processing is in stereo. On line 16, v is actually a pair of values, [left, right]. To process just the left channel, change line 14 to read

na[1, i - 1] = v[0]

Also, each_with_index starts at 0; using i-1 as the index puts the first sample at the end of the NArray. There is no observable effect from this since you're ALSO putting the first timestamp at the end of the NArray, and Fourier doesn't care about element order. :-)