LJ Rich finds out how technology is becoming more sophisticated in being able to distinguish songs from another and analyse their relationships.

In 2002, Click visited mobile start up firm Shazam, who was already offering a music recognition service.

The idea was to use a mobile to send a snippet of music which would be matched to a million songs on the firm’s servers at the time.

The user would then receive a text message with the track’s name.

Andrew Fisher, Shazam’s chief executive, explained that the different sound waves make identification possible.

"Every vocal performance has a different sound wave," he explained, "so it’s being able to very quickly look at that pattern and compare it against the eight million songs we have on the database, make the match and bring the result back to the consumer in around about five to 10 seconds."

‘In tune’

For users without the original music to hand, tech firm Melodis has created a desktop and mobile application called Midomi that aims to work out a tune from the crudest of renditions.

When someone sings into a phone, for instance, the program works by listening to the basic melody going in; it does not matter whether a person is singing, humming or even playing an instrument.

Melodis chief executive Keyvan Mohajer said the key and tempo is also irrelevant.

"But you do have to be in tune, and provide us enough information, to be able to match the song," he explained.

For people near tone deaf, their singing attempts can also be matched to those of other vocally challenged people.

"If a song is not in our database, [you can] sing or hum that song and save your voice, then your voice becomes a fingerprint," said Mr Mohajer.

"Next time somebody searches for a song, you match their voice to your voice to find the song."

Musican DNA

But humans are still much better at musical matching than computers – for instance we can recognise a song even if it is played in a different way.

Computers are not so sophisticated yet, but there are projects aiming to break music down into their constituent building blocks or "musical DNA".

Pandora’s Music Genome Project has created a collection of music analysis by analysing the musical qualities of songs.

These have their musical DNA charted by humans who identify "musical chromosomes" such as twangy guitars or interesting horn arrangements.

Machines are then able to provide the user with recommendations based on the results, and hopefully to their musical taste.

Charting origins

In contrast, mHashup uses a purely automated approach to discover musical relationships among tracks, displayed in a visual interface.

Unlike Midomi, it also matches timbre, or sound-type, as well as other more subtle qualities.

This sometimes produces results which seem quirky, like fitting Schubert to a Shakespearean sonnet. The system matches African tribal songs against Eastern European yodelling because of their similar melody.

However, the system is useful tool for charting the origins of music around the globe.

Michela Magas, mHashup’s designer, said she was approached by some Hollywood composers who were keen to test their compositions against large libraries.

"When they compose and they come up with something which they find is very appropriate for the movie… they cannot tell whether it is something they heard when they were young or whether it’s truly original – and this is a big deal," she said.

Of course, most human ears do all of this without needing the kind of brute force analysis and crowd sourcing that these projects rely on.

But machines are ever closer to developing a musical ear of their own.</p

This article is from the BBC News website. © British Broadcasting Corporation, The BBC is not responsible for the content of external internet sites.