The SemanticHIFI project, coordinated by the Paris-based music technology institute, Ircam, is unique. It represents a quantum leap in home music technology, in which access to musical content, and the ability to manipulate it, have hardly advanced since the days of the gramophone.
“Essentially, we are adding descriptions to musical content,” explains Ircam’s Hugues Vinet, the project coordinator. “This allows for more interaction with music, so users can do more than just passively listen. Actually, it’s about making our sophisticated software tools for professional musicians available to a broader public.”
These tools enable a wide variety of functions. Some address ways to browse the large number of recordings that now inhabit the average hard disk. “Browsing techniques for digital music were very basic,” explains Vinet. “You could only search ‘editorial’ information, such as titles. But SemanticHIFI will allow people to label and browse their own collections according to actual musical content, categorised as they see fit. “It’s not our object to define genres, but to let people define their own,” explains Vinet. “Then the system learns the definition criteria, and can label other titles accordingly.”
‘Browsing by example’ is another intriguing possibility – simply select the kind of music you want to hear, on the basis of features such as tempo or orchestration, and the programme comes up with a list of comparable pieces.
Naturally, exploring musical content in this way requires a mode of visualisation. “So we have developed a system that analyses the temporal structure of a piece of music and develops a graphical map or interface based on that,” explains Vinet. “So if you click on one of the elements in the graphical map, you go directly to that part of the music. What’s more, using this algorithm you can generate a musical summary as a new file – condensing a long piece into a much shorter one, but complete with all its variations. Then you can manipulate musical content via the summary and the graphical map.”
Another way of navigating through musical documents involves the ability to separate different instruments, using sound manipulation techniques that reproduce sounds in space. Here, SemanticHIFI challenges the usual recorded music model, which is undoubtedly polyphonic: “We have to persuade the music industry to evolve its production process by providing multitrack recordings,” says Vinet. Being able to separate the instruments allows the listener to arrange an orchestra in space, choosing where to place the violins, for example. It invites listeners and musicians to really understand the construction of a piece, and play along with it. “The system even includes simplified musical instruments that you can play with, using your voice,” Vinet explains.
SemanticHIFI’s system architecture has several components: a hi-fi box in the living room will house most of the capabilities. PC applications will enable more advanced functions, such as performance ones. Other capabilities are peer-to-peer file sharing – “In a non-copyright infringing way,” Vinet insists. “Users can share metadata – their indexing and manipulations – but not original tracks. The computer identifies the original behind the metadata, and if you don’t own it, will suggest you buy it. SemanticHIFI is therefore compatible with the commercial model.”
The project counts Berlin’s Native Instruments and the Sony European Technology Centre (Stuttgart) as its industrial partners. Sony handles the integration of the technologies into a box, which is the next step. “We’re two-thirds of the way there,” says Vinet. “All the technologies have been validated and the first application prototypes will be ready early in 2006, for a first trial at the Cité des Sciences in Paris.” He believes it is up to industry to decide the commercial future of the project: “The box itself may be a product, and parts of it may be adapted into mobiles or games – there are many possibilities,” he says.
But whatever form SemanticHIFI takes, one thing is for sure: listening to music will never be the same again.
SOURCE: AlphaGalileo