Mir

A couple of days ago, I read an interesting paper about a new AI algorithm that can summarize long texts. This is an attempt to solve the problem of tl;dr texts, meaning “too long, didn’t read”. The article reminded me that the same problem exists for music, in which case it would probably be tl;dl: “too long, didn’t listen”. I was interested in this topic back when I wrote my master’s thesis about short-term music recognition. One way to overcome the challenge of listening through full music tracks is by creating music “thumbnails”. That is, a compact representation of the most salient parts of the music in question. This is not a trivial task, of course, and lots of research have gone into it over the years. Strangely, though, I haven’t seen any of the many suggested algorithms implemented in any commercial service (so far). ...

I guess I was too much into NIME-organization back in March, to notice the launch of the The Million Song Dataset. It contains no audio, but 300 GB worth of metadata about 1 million popular music songs. This sounds like hours of great fun for music researchers around the world, and will probably also be a great resource for music students working on MIR-applications. I would also expect that it is possible to use this for a number of creative applications. ...

We organised a small workshop on Music Information Retrieval some weeks ago, and for that I carried out a small check for the most important MIR-topics using Google Scholar. I did this by first searching for the phrase “Music Information Retrieval”, which turned up 4670 hits. Then I started adding various other phrases, and the result was as follows: “Music Information Retrieval” + “…” 3730 - audio 1990 - MIDI 544 - action 485 - motion 328 - gesture 85 - movement action gesture 45 - “motion capture” 21 - “body movement” Quite clearly, audio-based methods seems to dominate the MIR literature, while symbolic representations (e.g. MIDI) is also quite high on the list. Words like action, motion and gesture turn up some hits, but these words are quite general, and may refer to many different things. It is interesting to notice how few hits there are for “body movement” and “motion capture”. Quite clearly there is room for a lot more MIR research from an embodied perspective.

CMMR, Pisa, Italy 26-28 September 2005 This was a rather small conference, with only about 40 participants, organised at the CNR in lovely Pisa. The topics presented were varied, but here, as in most other computer music conferences these days, there were a high percentage of music information retrieval presentations. I was there to present a short paper on building low-cost music controllers from hacked gamepads and homemade sensors, something I worked on while at McGill in the spring. A summary of things I found interesting: ...

My master’s thesis (cand.philol.) has now been submitted to the University of Oslo: How Do We Recognize a Song in One Second? the Importance of Salience and Sound in Music Perception. Abstract This project started with the observation that we manage to recognize a song by listening to only a second of it. What perceptual and musical features make this possible, and can such features be used in music analysis and music information retrieval? These questions can be broken down to two main problems: a) segregation of sensory input and b) recognition of musical features. ...