please visit makeyourselftransparent.tumblr.com
The relationship between the music and video in this piece is determined by a simple learning algorithm. The music is analyzed by extracting mel frequency cepstral coefficients (MFCCs) in realtime, which could be likened to a spectral fingerprint. I selected certain MFCC frames to train the learning algorithm, which then computes the distance between them and the new frames. The gap between the familiar and unfamiliar spectral fingerprints is constantly changing, which causes constant changes in the processing done to the video material.