First problem we have to face is how to handle output differences between suivimidi and suiviaudio: In mididetector detection algorithm, we use the current state of the follower to detect errors in alignment. We can't do the same thing for the moment with audiodetector, because suiviaudio doesn't output this current state.
We have so 2 possibilities:
- find another way to detect errors,
- change outputs of object suiviaudio.
First solution would be a good way to find improvements for the program (by combinating 2 algorithms for example), but is also time demanding. Second solution would be the most logical (it is not very difficult to add an output to object suiviaudio), but requires to be done by jMax and Score Following Developpers.
Second problem will come with all parameters we will have to determine before we are able to detect errors: for example, start time of a note will have to be checked out by looking at energy and spectrum parameters, i.e. with many calculations.