I will guess this tune in three notes!

    So can answer now a computer algorithm specially trained by a team of researchers, capable of determining whether a melody belongs to different periods (Baroque, Classicism, Romanticism and Postromantism) in just three notes from any part of the musical composition.

    True, the system itself cannot be identified by three notes. But this is not required - the developers set themselves the task of analyzing the evolution of music over a period of 200 years (1730-1930), and teaching the computer system to correctly determine whether a melody belongs to one of the above periods / styles.

    The essence of the study

    The developers tried to highlight separate patterns for each period / style, including the texture of the melody, harmony, rhythm and melody.

    The training of the computer system was carried out on the basis of the well-known Peachnote service , to which more than 20 thousand musical works of the styles described above were previously added. The machine was able to distinguish certain patterns of musical works by semitones or notes. After conducting cluster analysis, the results were summarized in a single graph, which shows the stages of the evolution of musical styles, as well as transitions between styles.

    Here, the transition between the various musical styles of the past is shown quite accurately.

    The computer helped to identify the unique features of each of the analyzed musical styles, and a clear connection was determined between the distribution of intervals in a particular musical composition and a person's ability to predict the development of sound composition. By the way, a popular theory claims that a person is able to predict the development of musical composition in just three notes. As we see, now the computer system can do the same, after appropriate training.

    As far as one can judge, a person perceives music depending on the subconscious “expectation” of the distribution of musical intervals in each particular composition. The main criterion is the distance between two consecutive notes. Accordingly, by three notes a person (and now a computer) can predict how a musical composition will develop.

    Why all this?

    In principle, scientists / engineers often learn something / build theories for the sake of pure science, without application to practice. In this case, the authors of the project hope to apply their achievements to highlight patterns of human speech. This is done for the early diagnosis of Parkinson's disease, for example, when everything is fine with motor activity, but the voice is already changing, and only computer analysis can show this, revealing the disease at an early stage.

    In addition to Parkinson's disease, it is possible to distinguish the characteristic patterns of human voice for other diseases, including mental ones.

    Probably in the near future, such diseases can be diagnosed without problems, with verbal communication with a person (for example, by phone).

    Also popular now: