I have become interested in the MIDI input again, following another flurry of emails on the LilyPond mailing list. I've been experimenting with a neural network (the package is called Fann) and have got it distinguishing a sequence of four quarter notes from a sequence of two dotted-quarter-eighth notes ... It doesn't seem to be quite the right tool, as it is a classifier with a fixed number of input and output neurons. In the case of MIDI-in we have a stream on input events, it's more like the speech recognition problem. I haven't been able to understand yet what sort of approach is used for speech recognition, but I thought I might try the Fann classifier on classifying the duration of single notes by giving it the durations of, say, two notes on either side as context. If you or anyone else on the list has some ideas on this I'd like to hear.
Richard _______________________________________________ Denemo-devel mailing list [email protected] https://lists.gnu.org/mailman/listinfo/denemo-devel
