Hledání slovníku pro audiosignály

Abstract

Tato prace by mela odpovedet z ceho se sklada audio signal. Zabyva se tedy ucenim slovniku pro audio signaly na zaklade ridkych reprezentaci. Jsou zde popsany jednotlive algoritmy, ktere jsou dulezite k vytvoreni prizpusobenych slovniku. Dale v praci nalezneme srovnani reprezentace signalu pomoci fourierovy transformace a natrenovaneho slovniku. Je zde popspana tvorba databaze hudebnich zvuku.

This thesis should answer what constitutes audio signal. It deals with the dictionary learning based on sparse representations of signals. This thesis describing algorithms, which are important for the creation of learning dictionaries. We find a comparison of signal representation using Fourier transform and learned vocabulary. It describes the creation of a database of musical sounds.