FFT power spectrum visualizer

A high-quality demonstration of the application can be found at

• https://www.youtube.com/watch?v=AaJ25F8pjOw
• https://www.youtube.com/watch?v=HsuVb4TDCqU

This program utilized discrete Fourier transformation to obtain the time-frequency-domain of a audio file. The result can be explained as following:

• The top part of the program, we can see the raw time-amplitude data as it can be found in any WAV file. While this is the intuitive way for how we think of data, it is not very informative.
• The bottom part of the program is the output of the algorithm, combined with some waterfall effect. This parts display the heart of DFT or Fourier Transformation in general, which allows us to decompose the data into their frequencies and how much each frequency contributes to the final sound. This kind of data has many nice attributes to which allow us to deploy many other tools from signal processing, machine learning and pattern recognition to wrangle data in a more effective way.

This program is on the other hand just a nice artistic demonstration of the algorithm, notice how much better we can see which piano key is being pressed, which is almost impossible to achieve with time-amplitude.

Build and run it yourself (C++11 / SFML 2.5)

Install the dependencies

sudo apt-get install libsfml-dev

Clone the repository

git clone https://github.com/longmakesstuff/FFT-Power-Spectrum.git --depth 1

Compile

cd FFT-Power-Spectrum
mkdir build
cd build
cmake ..
make -j8

And play any WAV song you want with

./bin/main [song-path.wav]

References

• Louisa Renz's bachelor thesis - University Tübingen
• RosettaCode's Cooley-Tukey FFT
• Hann's function
• Fourier transform for dummies