Uma abordagem baseada em otimização para geração de licks guitarra

Abstract

The present work describes the operation of a computational approach that, through the use of exact and heuristic methods, performs the construction of melodic fragments, which are represented, more speci cally, by guitar licks. This type of application is part of the eld of Algorithmic Composition, which is the creation of music with the aid of algorithms. For this purpose, a database of smaller melodic fragments entitled \sublicks" is used. These fragments are the most elemental portion of a guitar lick, which is the musical notes. Given this, a set of rules are established to score transitions between sublicks, so we can determine a cost matrix to assist in their concatenation, resulting in a complete lick. Such concatenation is performed by a mathematical model, which still takes into account certain inputs, such as the amount of musical notes and guitar techniques to be used in the construction of lick. Additionally, a probabilistic Markov Chain-based model was necessary to construct a di erent type of lick, called a turnaround lick, due to its peculiar characteristics. To validate a proposed tool, melodies resulting from it were subjected to human evaluation with a selected group of 60 people with di erent levels of musical knowledge. Overall, the results suggested that there was a good acceptance by the listeners, noting the pleasantness of listening to these melodies.