I presented a poster on Joshua Steele at the Digital Music Research Network (DMRN) workshop this week. More on the poster in another post but the day did provide a range of talks.
Much enjoyed the keynote by Augusto Sardi about capturing and rendering spatial audio. It reflected on techniques from computer vision and the notions of ray space and plenacoustics but worth the price of entry… I’ve made a note that it echoes Kittler and storage but it may the way the plenacoustic microphone and systems can reproduce audio streams.
Bob Sturm’s talk on the use of Machine Learning to learn from and create folk music was exciting on so many levels. I had heard of the work from Thor Magnusson and really wanted to catch it. Sturm reflected on Wagstaff’s Machine Learning that Matters and its suggestions towards the Machine Learning’s relations with a wider world. Apart from Duncan Williams’s talk on bio-responsive control, it was one of the only papers that really dealt with people.
I had seen Luca Turchet discuss his augmented mandolin at Audio Mostly and this was extended into the Internet of Musical Things.
The day was heavily machine learning oriented and how it can help with digital musicology. I was wondering where the human element was in the process and there did seem (apart from Bob Sturm) to have surprisingly little reflection on the way knowledge was being processed and created. Rather, there was a more computer science approach to optimisation. In the final talk, Peter MC Harrison on a statistical learning model for perception, there was mention of abstract symbols to represent music, itself a sequence, for the ML process. He did suggest a bias towards classical music in the psychology tests, raising questions if focusing on jazz or pop.
It was wonderful to catch up with people there and some interesting ideas are bubbling.