A novel MIR framework and application with automatic voice processing, database construction and fuzzy matching

Abstract

In this paper, we present a music information retrieval system which enables users to retrieve music by vocal query. Three essential components are query processing, database construction by MIDI and an approximate search engine. For query processing, we have achieved a real-time and robust voice-to-melody converter. For database construction, proposed MIDI analysis methods to obtain music melody features from MIDI files automatically. In order to match query with melodies in database, we extend an existing search engine into a fast approximate melodic matching engine. We have carried out extensive experiments on the prototype system to evaluate the performance. The results show that the proposed three components are achieving good performance.