We introduce a tempo-invariant feature named ``rhythm vector,'' since
the tempo of the input data is not given in advance and it may vary
throughout the data. From our assumption that the tempo is constant or
changes slowly, the proportion of consecutive note lengths is nearly
independent from tempo according to Eq.1).
Therefore, we introduce rhythm vector as follows:
Here, probabilistic distribution
is assumed to follow the
, whose parameters,
and covariance matrix
, can be
obtained through statistical training using human-performed MIDI data.