Korvapuusti is a psychoacoustic model that quantifies the human observable difference in two audio signals in the proximity of just-noticeable-differences.
In this project we study the psychological and physiological responses associated with sound, to create a new more accurate model for measuring human-subjective similarity between sounds. The main focus will be on just-noticeable-difference to get most significant benefits in high quality audio compression. The main goals of the project is to further both existing and new practical audio (and video containing audio) compression, and also be able to plug in the resulting psychoacoustic similarity measure into audio related machine learning models.