As most of us know, we can express pure music tones such as the A above middle C on the piano by simply writing that it is a note of frequency 440 hertz (Hz) with wavelength of 78.4 centimeters (cm). What if, instead, we wanted to mathematically write down the music-playing of different musicians in a way that not only reflected the different frequencies and wavelengths of notes, but also encapsulated the unique expressions of their playing? The authors use trend-based modeling to address this music and computer science challenge.
The authors extract descriptors and then perform a qualitative analysis of detected trends. The modeling captures different violinists’ expressive tendencies. They characterized the violinists by a set of frequency distributions and melodic patterns. Their several experiments report good results. Also, they based their work on some professional violinists’ commercial recordings.
The authors’ continued research will involve expanding their constraints to deal with polyphonic recordings and using fuzzy theory to improve results. This is a very interesting data analysis problem with tangible results.