diff options
| -rw-r--r-- | README.md | 4 |
1 files changed, 1 insertions, 3 deletions
@@ -5,8 +5,6 @@ This repository contains the realization of the [second task](http://home.agh.ed Makefile README.md - make_results.sh - template.gnuplot src/ fourier.F90 f1_dft.gnuplot @@ -82,4 +80,4 @@ And the plot of it's dft (with _x_ axis range of [0; 150] and logarithmic _y_ sc We can see, that there's a single pike at frequency 1 (corresponding to our main cosinus signal) and all other values are low and seemingly random. The last part of the task was to remove all values of the dft with absolute value smaller than 50 and perform the inversed Fourier transform. This zeroed almost all of the 'random' part of the dft. Since the __noise__ in the dft was the result of the __noise__ in the original function, the result of the inversed Fourier transform was an almost perfect cosinus as can be seen below  -This worked, because the amplitude of the noise was small and so the amplitude of frequency components resulting from it was small (almost all absolute values of dft for frequencies other than 1 were 0). In case of stronger noise or a cap smaller than 50 this could possibly not work as well. Nevertheless, this (fast) Fourier transform example also presents one of the method used used for filtering periodical signals from noise. Other ways of filtering dft values also exist (e.g. convoluting with a predefined filter-function).
\ No newline at end of file +This worked, because the amplitude of the noise was small and so the amplitude of frequency components resulting from it was small (almost all absolute values of dft for frequencies other than 1 were smaller than 50). In case of stronger noise or a cap smaller than 50 this could possibly not work as well. Nevertheless, this (fast) Fourier transform example also presents one of the method used used for filtering periodical signals from noise. Other ways of filtering dft values also exist (e.g. convoluting with a predefined filter-function).
\ No newline at end of file |