Some New Methods of Music Synthesis

Abstract

There are two distinct sections to this thesis. The first section discusses music composition, shows why it is a useful domain for Artificial Intelligence research and presents a set of "Design Rules" that facilitate research in the field of tonal music composition. It begins with a short chapter presenting a subset of music theory. This chapter assumes no prior knowledge of the subject, it completely defines all terms used in the thesis, and is geared particiularly toward those unfamiliar with music, those unwilling to learn standard music notation and those interested in Artificial Intelligence research. Next, (using the terms defined in the thesis), a context sensitive generative grammar for producing pitch progressions in the major mode is introduced. It is seen that the grammar can be made context free by switching between two interpretations of the input string. A mechanism for switching from one interpretation to another when parsing sentences generated from this grammar is described. It is shown that a model of music composition, perception, and improvisation fits within the framework of the grammar. This multiple view model and switching mechanism can be interpreted as a primitive "frame." The section section describes some of the problems and issues encountered while designing the initial hardware for the Music Aided Cognition Project at MIT. All of the developed hardware permits computer control, performance and recording of music in real time. The first chapter in this section discusses a machine called the Inexpensive Synthesizer/Recorder. It capable of synthesizing 14 square wave voices, each voice having a range of 7 octaves, with each octave having 12 bits of frequency control. Its purpose it to allow the user to record key depression times, key release times and key impact velocities when playing a keyboard piece. Its primary constraint was low cost, allowing many copies to be made. Its microprocessor interface allows it to be easily controlled by many different means, including home computers. The complete schematics for the synthesizer and the controller are provided as an appendix. The next chapter discusses an oscillator which synthesizes sound using 32 sine or 8 FM waveforms. The machine can be easily expanded to produce 256 sine voices and 64 (or more) FM voices. All since waveforms in both types of synthesis are weighted with two independent coefficients. Micropogrammable firmware allows one to produce sound by a limited number of methods other than sine summation or FM synthesis.