Generation of jazz improvisations in MATLAB

Abstract

A menudo se dice que "el genio no nace, se hace". Sin embargo, a menudo se considera que los procesos creativos requieren alguna forma de don innato, a veces místico en su naturaleza. La música es una de estas actividades, donde incluso aquellos que la practican no están seguros de donde vienen sus composiciones a pesar de haber empleado años de trabajo en convertirse en maestros de sus instrumentos. Uno puede afirmar que nuestra habilidad para adaptar rápidamente nuestras experiencias pasadas a nuevos contextos es una de las características que define la inteligencia humana. En nuestra vida aparecen situaciones que no nos son familiares donde tenemos que recurrir al juicio, aplicando el curso de acción que consideremos más apropiado y basando nuestra decisión en sucesos previos y teniendo en cuenta los nuevos factores que hacen que las circunstancias sean diferentes con el fin de responder de manera correcta. Tal problema emerge también en la música, en particular en el género del jazz, donde la improvisación es un rasgo clave. No importa lo complicado que resulte ser un solo, la improvisación es una habilidad que puede ser aprendida. Sin embargo, no está claro cómo se adquiere en realidad esta habilidad. Una de las teorías imperantes sugiere que la repetición continuada de material similar lleva a la aparición de patrones en la obra del solista, que a menudo abarca años de su carrera. Nos basaremos en esta teoría y la exploraremos en detalle para a continuación describir y ampliar la implementación de un algoritmo para la generación de solos a partir del corpus del saxofonista Charlie Parker. El algoritmo empleará cadenas de Markov para producir una secuencia de notas que guardarán una semejanza al material de Parker debido a la aparición de motivos musicales recurrentes. Además se ofrecerá al usuario la posibilidad de insertar sus propias progresiones de acordes para restringir la salida del programa a través de una interfaz de usuario. Esperamos que esto aporte algo de luz al debate de la improvisación. La improvisación no es solamente un problema musical fascinante, sino también una parte de nuestra vida cotidiana. Un mejor entendimiento de la improvisación conlleva un mejor entendimiento de la mente humana, y todo paso hacia la comprensión de nuestro propio pensamiento nos ayuda a encontrar la respuesta a una de las preguntas más importantes de la humanidad:¿cómo funciona nuestro cerebro? It is often said that \a genius is made, not born." However, creative processes are often considered as acts that require some kind of innate gift, sometimes mystical in its nature. Not surprisingly, music is one of those activities, where even practitioners themselves are unsure of where their compositions come from, and this is despite their spending years working to become masters of their instruments. One can afirrm that our ability to quickly adapt our past experiences for new contexts is one of the traits that defines human intelligence. Unfamiliar situations arise where we have to draw upon our judgment, applying the course of action that we deem most appropriate basing our decision on previous occurrences and taking into account new factors that make the circumstances different in order to respond in a fitting way. Such a problem also emerges in music, particularly in the jazz genre, where improvisation is a core feature. Improvisation emerges from combining creativity and adaptation. It may well be defined as an immediate musical composition. It is one of the most difficult abilities to master as it demands a complete command of the instrument as well as exceptional creativity and musical thought. It forces the artist to adapt her playing at an extraordinarily fast pace, while they move within the confines of the song that are being built, in real time, by the rest of the band. In jazz, this complexity is aggravated by the unconventional chord progressions and frequent key changes (unconventional understood as deviation from the common practice). No matter how challenging and intricate a solo happens to be, improvisation is a skill that can be learned. Proficient jazz artists such as John Coltrane or Dizzy Gillespie would not have felt the need to practise scales everyday for decades if that were not the case. Nevertheless, it is unclear how this skill is actually acquired. Some experts contend that solos are the result of implicit musical rules that dictate the sequence of notes that shapes the output. Others, however, argue that the continued repetition of similar material leads to the appearance of patterns in the soloist's work, sometimes spanning years of their career. This document will touch upon previous research that explores both options, as well as go through some examples of commercial and academical software that tackle the topic of improvisation from different standpoints. We will focus on the former theory and will go into detail about the implementation of a pattern-based algorithm for the generation of solos based on the works of jazz saxophonist Charlie Parker, whose transcribed works are easily available on the internet and other sources. This algorithm, originally presented in Norgaard et al. (2013), will employ Markov chains, a concept that will be thoroughly described throughout the document, to generate a sequence of notes that bear a resemblance to Parker's material, partially through the appearance of reoccurring musical motifs. In addition, our improved program will offer the possibility of inputting chord progressions to constrain the output of the program through a user interface. We hope this work will help shed some light on the ongoing debate between the aforementioned schools of thought. Improvisation is not just a fascinating musical problem, but also a part of our everyday lives. A better understanding of improvisation is a better understanding of the human mind, and every step towards comprehending our own thinking helps us find the answer one of the most important questions for mankind: how does our brain work?