German vs. Austrian folk song classification

Computerized analysis and classification of folk songs receives increased attention in recent years. In this paper, we demonstrate a two-case German and Austrian folk song classification using the musical feature density map (MFDMap) as the musical features representation and the finite impulse response extreme learning machine (FIR-ELM) as the machine classifier. Fifteen different MFDMaps are designed to study the music properties that aid in characterizing the differences. Our simulations show that the FIR-ELM classifier can achieve 83% classification accuracy using the MFDMap with interval, duration and duration ratio features.