Core processing engine of cava.
cava core works as a wrapper around fftw. While fftw does the job with the discrete Fourier transform which transforms the audio from the time domain (series of samples) to the frequency domain (series of frequencies), the raw output data from fftw will not look very nice when used directly in a visualization. The cavacore engine does several things to improve the look of the visualization:
Theoretically the human ear can hear frequencies up to 20kHz, but the information above 10k is hard to separate from eachother and does not give much feedback by itself when visualizing audio. Similarly audio down to 20Hz can be heard, but the frequencies below 50Hz does not directly provide anything to the visualizing of the audio.
The human ear hears different frequencies logarithmically, so notes that are 2x and 4x higher than eachother, will be preceived as beeing a fixed amount higher than eachother. Therefore cavacore outputs the frequencies logarithmically. Cavacore will also group the output in the desired number of samples.
the raw output of fftw is very noisy, cavacore processes the output signal in two ways to provide a smooth output:
- an integral filter calculates a weighted average of the last values
- fall off filter, when values is lower than last value it uses a fall down effect instead of the new value
This feature can be adjusted.
The range of an input signal can vary a lot. cavacore can keep the output signal within range in real-time. This feature can be disabled.
See cavacore.h for documentation and the cavacore_test.c application for how to use.