| Due to the large amount of musical data available on the internet in recent years, efficient and intuitive methods are required for searching the musical data. Musical search services, such as the iTunes provides, support querying capabilities on the basis of metadata tags (title, artist, etc) associated with the musical data. The natural way of searching musical data is to search by its content rather than secondary features like title, genre etc, because the content is usually more memorable. In this research, content-based music retrieval is performed on a polyphonic MIDI music database where the query is a hummed tune. Two approximate string matching algorithms, LCTS and Myers algorithms are modified, applied to the problem, and retrieval performance is calculated. Response times of the algorithms are calculated by altering the values of some of the interesting parameters such as the query length, degree of polyphony and size of the database. |
