For use with EqualizerAPO.

Generates an EqualizerAPO Graphic EQ preset from 32-bit floating point raw audio files of pink noise to adjust output frequency response, such that it closely approximates the a-weighing curve. Removing the a-weighting can approximate a flat response, although this is not ideal for music.

For use in WSL2, simply run python3 process.py file.raw | clip.exe to copy a preset to your clipboard.

Power spectral density is computed over overlapping segments of the signal using SciPy's scipy.signal.welch function.

The measured spectral density is normalized to fit a target 1/f curve. Then, the target density is divided by the measured power density pointwise at each measured frequency, and the the ratios are converted to power decibels. The resulting decibels represent an approximate gain to apply to each frequency band to flatten the response on output.

The resulting curve is then a-weighted.