It is capable of analyzing an audio fragment, separating the most prominent frequencies, analyzing the sound and creating a sound map that is transformed into a numerical value
Professors from the International Universities of La Rioja and Valencia have created "SpectroMap", an algorithm that
Accurately measures similarities between songswith which it can become an effective tool in the fight against
plagiarism in the music industry and, at the same time, it can have creative applications, to build melodies.
This was explained to Efe by Brian Martínez, professor of Music and Musical Research at the International University of La Rioja (UNIR), who is developing this project together with researchers Aarón López-García and Vicente Liern, both from the University of Valencia, specialists in Scientific Computing and in Mathematics for Business, respectively.
They have already presented their project and the first design of the algorithm at an international congress on music and mathematics, although «SpectroMap» still needs more development, the UNIR professor has clarified.
The creation of this algorithm is based on the idea that
«music is creativity, yes, but also mathematics and rationality»he assures and, in fact, «there have been musical creations for centuries that have used mathematical procedures».
With that idea, more than three years ago they advanced in the project of measuring "objectively" the similarities between songs "something that seems totally subjective"; Thus, they advanced to create an algorithm that measures patterns within songs, classifies styles or measures melodic similarities, among other characteristics.
"The basic concept is to create an objective procedure to compare songs and what degree of similarity they have," explains Martínez.
There is already technology that goes along this line, even in popular use, he acknowledges, but they have sought to develop a more complete algorithm, which is
capable of analyzing an audio fragment, separate the most prominent frequencies, analyze the soundfilter all the information provided and be able to create "a sound map" that is transformed into "a numerical value".
In the initial presentation of the project they have already provided practical work with three songs, by Adele, Bruno Mars and John Legend, although they have only worked with versions of their songs that circulate on the internet performed by other artists.