Cognitive Musicology: Segmentation and Pattern Recognition

This project originally began with financing from the CNRS program (Action Concertée Incitative) "Complex Systems in Humanities and Social Sciences". The project continues today in collaboration with the Université de Strasbourg and the
CIRMMT.

The central idea of this multidisciplinary project is to design a model that is both computational and cognitive of the mechanisms of perception and comprehension of musical patterns at work in improvised music from the Mediterranean and in Western jazz music looking specifically at the interaction between the process of segmentation and that of the recognition of a work when listening from different points of view - perception, musical analysis, and computer modeling.

This question is traditionally found in the field of objects that are complex in terms of their representations of knowledge schemas, of hierarchical segmentation processes, of interaction between ascending and descending processes, and above all for everything that concerns music, the problems of application of these operations in real-time when listening to music. This research is centered on the modeling via a computer system musical works with analytical and perceptive hypotheses, in a double exchange with experimentation: provide an analysis-synthesis setting favorable to the validation of the hypotheses and take the pertinent indications that are likely to lead to the improvement of the model. Such an objective requires close collaboration between all involved disciplines. Even if each domain brings its own methodology, we favor discriminating interaction between the different steps of analysis, modeling, and experimentation.

Transcultural experiments with acculturated subjects have already been carried out in different styles are addressed in this study and the results have been compared with those provided by a computer model, placing an important concept in the forefront - that of the mother melody or the archetypical pattern that spreads to the specific melodies. The recognition of these archetypes is distinct for each of the two populations and it seems that computer modeling should take this aspect into account in order to reach a convincing level of analysis.

Partners: Université de Strasbourg, CIRMMT McGill University

Participants