Temperley D. and D. Sleator, “Modeling Meter and Harmony: A Preference-Rule Approach,” Computer Music Journal, 23(1), 10–27, Spring 1999.

Abstract

Computational music analysis is an important project for a number of reasons. From a psychological point of view, systems for performing musical processes may shed light on how human listeners perform such processes, just as computational work in vision and problem solving has led to important insights in those domains. While recent research has revealed much about listeners’ mental representations of music, there is much to be learned in this area. There are practical applications to automatic music analysis as well. Many kinds of music-related tasks that computers might perform—for example, producing a score from raw pitch data, detecting probable errors in a score, searching a melodic database for matches to a query, generating an accompaniment for a melody, or performing statistical analysis of musical styles for the purposes of automatic composition—require an ability to process music and extract various kinds of structure and information from it. Extracting this information proves to be a highly complex task.
In this article, we present a computational system for analyzing metrical and harmonic structure. The system is designed for Western tonal music, particularly art music of the “common-practice” period.