|
Abstract : |
Two systems are reviewed than perform automatic music transcription. The first perform monophonic transcription using an autocorrelation pitch tracker. The algorithm takes advantage of some heuristic parameters related to the similarity between image and sound in the collector. The detection is correct between notes B1 to E6 and further timbre analysis will provide the necessary parameters to reproduce a similar copy of the original sound. The second system is able to analyse simple polyphonic tracks. It is composed of a blackboard system, receiving its input from a segmentation routine in the form of an averaged STFT matrix. The blackboard contents an hypotheses database, an scheduler and knowledge sources, one of which is a neural network chord recogniser with the ability to reconfigure the operation of the system, allowing it to output more than one note hypothesis at the time. Some examples are provided to illustrate the performance and the weaknesses of the current implementation. Next steps for further development are defined. 1., |
