作者: Elias Pampalk , Masataka Goto
DOI:
关键词:
摘要: In this paper we present MusicRainbow which is a simple interface for discovering artists where colors encode different types of music. based on new audiobased approach to compute artist similarity. This scores 15 percentage points higher in genre classification task than the similarity computed track level. Using traveling salesman algorithm, similar are mapped near each other circular rainbow. Furthermore, combining audio-based information with from web. particular, label rainbow and summarize words extracted web pages related artists. We use vocabularies hierarchical levels heuristics select most descriptive labels. conclude discussion results. The first impressions very promising.