Demystifying the Fourier Transform: The Intuition

참고 : https://www.youtube.com/watch?v=XQ45IgG6rJ4&list=PL-wATfeyAMNqIee7cH3q1bh4QJFAaeNv0&index=1


1. Intuition FT

Decompose sound into multiple frequency components

( frome TIME domain to FREQUENCY domain )

figure2


Summary

Compare signal with sinusoids of various freuqencies

  • for each frequency, opbtain (1) magnitude & (2) phase
  • high magnitude = high similarity btw signal & sinusoid


Sine wave

\(\sin (2 \pi \cdot(f t-\varphi))\).

  • \(f\) : frequency
  • \(\varphi\) : phase


2. Procedure of FT

Step 1) Choose a frequency

Step 2) Optimize phase \(\varphi\)

  • that maximizes the similarity with the signal

Step 3) Calculate magnitude

  • magnitude = similarity btw signal & sinusoid of chosen frequency

\(\rightarrow\) REPEAT 1~3 for all possible frequencies ( inifinte..? )


Optimizing a phase

\(\varphi_f=\operatorname{argmax}_{\varphi \in[0,1)}\left(\int s(t) \cdot \sin (2 \pi \cdot(f t-\varphi)) \cdot d t\right)\).

  • \(s(t) \cdot \sin (2 \pi \cdot(f t-\varphi))\) : multiplying signal & sinusoid

  • \(\operatorname{argmax}_{\varphi \in[0,1)}\) : selecting the phase that maximizes the area


Calculate the (largest possible) area

( with the chosen best \(\varphi\) )

\(d_f=\max _{\varphi \in[0,1)}\left(\int s(t) \cdot \sin (2 \pi \cdot(f t-\varphi)) \cdot d t\right)\).


BUT …. infinite range!

  • \(t \in R\).
  • \(f \in R\).


3. Inverse Fourier Transform

Reconstructing the original signal

figure2

Categories: ,

Updated: