This short paper presents a novel interaction technique for smartphones that allows users to browse music stored on mobile devices in a game-like way. Unlike music retrieval methods based on metadata text search, the nepDroid mobile music player uses clustering and visualization techniques to display relevant music information. The system positions a small Droid icon at the center of the interface, which shows music collections retrieved from the mobile phone’s memory in a color landscape format. “Island” visualizations represent clusters of similar types of music, such as hard rock. Notably, audio features such as mel-frequency cepstral coefficients (MFCCs), which “describe timbre,” and fluctuation patterns (FPs), which “reflect rhythmic properties,” are extracted, and a clustering algorithm is used to display the distribution of music items in a 3D space. Clusters are visualized via a smoothed data histogram (SDH), and high values of music density are mapped to color values.
When the main icon navigates close to a song, nepDroid automatically plays it. From the user experience point of view, this represents a very promising approach to music navigation and visualization. Due to the high complexity of the system, the visualization of a library of a few hundred songs requires high computational hardware resources, which are limited even on latest-generation mobile phones. However, the system is a viable approach for small-scale computation and visualization.