An Australian computer scientist says that the next generation of search engines will be able to retrieve music for you when you sing a song to your computer. She said that 'in the next three or four years it should be on the computer of everyone who is a music fanatic.' There are still problems to solve, such as the diversity of music genres and 'the effect interference from things such as sound quality and environmental noise have on the ability of the program to detect notes.' And she knows what she's talking about: she's also an accomplished musician who is publishing her second album.
This project on a specialized search engine has been led by Alexandra L. Uitdenbogerd, a lecturer of the department of Computer Science at the RMIT University, Melbourne, Australia. She's also one of the researchers in the Search Engine Group. Moreover, she's a real musician who's using the simpler name of Sandra Bogerd for her activities outside RMIT. You can see a picture of her on the left. Sandra's music covers a wide variety of styles, with a tendency for middle eastern and Indian-influenced pop and instrumental works. She is currently recording her second album.
But it's time to look at the ABC Science Online article to get some details about her music search engine. "Uitdenbogerd says one form of retrieving audio by singing will involve users calling up a specific website then singing a tune or lyrics into a computer microphone to submit their query. The computer will then search the website's database to retrieve a menu of digital files, which the user can then choose from to download. But the quality of your voice will affect the search."
She just means that some people will have to sing for a longer time than others, but that ultimately the system will find the correct song even if you're an terrible singer and if some environmental noise alters the ability of the program to detect notes. "These outside influences can affect the frequency of the notes the program is trying to interpret. 'Audio is just a wave form that goes up and down and doesn't bear much resemblance to how we perceive [music] as notes,' she says."
One of the latest research works of Uitdenbogerd has been published in the INFORMS Journal on Computing under the name "Methodologies for Evaluation of Note-Based Music-Retrieval Systems" (Vol. 18, No. 3, Summer 2006, pp. 339-347). Here is a link to the abstract. "There have been many proposed music-retrieval systems, based on a variety of principles. How the effectiveness of these systems compares is not clear. The evaluation of some systems has been informal, without the rigor applied in other areas of information retrieval, and comparison of systems is difficult because of the lack of a common data set, queries, or relevance judgments. In this paper we explain how we collected artificial and expert music queries and name-based relevance judgments, and describe software we developed for collection of manual relevance judgments. Together with a collection of downloaded musical instrument digital interface (MIDI) files, these sets of queries and relevance judgments provide valuable tools for measuring music-retrieval systems. As an example of the value of these tools, we use them to compare the effect of using the expert queries and manual judgments to that of the artificial queries and manual judgments used in our earlier experiments."
But if you really want more information, you can read her PhD thesis submitted for the degree of Doctor in Philosophy in 2002, "Music Information Retrieval Technology" (PDF format, 186 pages).
A final question remains: will the dominant search engines of the world will buy this technology?
Sources: Dani Cooper, ABC Science Online, July 25, 2007; and various websites
You'll find related stories by following the links below.