Search by query is a common service offered in libraries, information databases, and the Internet, but these are currently limited to word based queries. A method, known as "Query By Humming," has been proposed for using melody based queries. The current widespread use of word-based searches suggests that a query by humming search mechanism may be a useful method for searching for songs or tunes in the future. This project tried to determine whether music can be simply and effectively analyzed and stored so that it can be accurately retrieved by someone humming the tune.
A musical encoding method was created to allow for a practical method of storing musical data for search. The encoding method uses the difference in chromatic steps between notes to represent songs. A library of music was then created to provide data to search through. In order to find a specific piece of music that a user has hummed, a Java computer program was written to search the music library for a set of notes.
Data was collected by humming parts of songs and recording them as computer files. Using a phonetic analyzer, the average pitch of each note was determined. After converting the difference in pitch between each hummed note to a number of chromatic steps, this string of relative changes in notes was used as the search criteria. The searches returned the correct song 75% of the time.