Using Expressive Trends For Identifying Violin Performers

This paper presents a new approach for identifying professional performers in commercial recordings. We propose a Trend-based model that, analyzing the way Narmour’s Implication- Realization patterns are played, is able to characterize performers. Concretely, starting from automatically extracted descriptors provided by state-of-the-art extraction tools, the system performs a mapping to a set of qualitative behavior shapes and constructs a collection of frequency distributions for each descriptor. Experiments were conducted in a data-set of violin recordings from 23 different performers. Reported results show that our approach is able to achieve high identification rates.