Researched the use of metaphor-based metaheuristic feature selection algorithms in music genre classification. Experiments were ran comparing:
- PCA
- ReliefF-SFS
- Self Adaptive Harmony Search (SAHS)
- Binary Cuckoo Search (BCS)
- Binary Dragonfly Algorithm (BDFA)
SAHS, BCS and BDFA were then used to build features sets for binary classifiers e.g. rock vs pop, reggage vs classical.
Rhythm patterns extracted using https://github.com/tuwien-musicir/rp_extract