International Symposium on Nonlinear Theory and its Applications
Music Similarity Measure using a Nonlinear Time Series Analysis Method
Ken’ichi Sawai, Yoshito Hirata, Kazuyuki Aihara,
PDF download (131.9KB)
This paper describes how to apply a nonlinear time series analysis method to musical scores. The technique is called cross recurrence plot. This can measure the similarity between two time series data. In our study, we use symbolic music information as obtained from musical scores. By regarding the symbolic information as time sequences, we can consider applying cross recurrence plots.
When we use cross recurrence plots, it is necessary to define the distance between two time points. If the data to be analyzed were numerical, the distances could be easily defined as Euclidean distances or infinity norms, for example. However, since musical data are the objects of the analysis, a new distance function is needed. Therefore, we propose a kind of edit distances by extending the distance for neuronal impulse trains. By using cross recurrence plots, we can quantify the atmospheres of musical pieces.