Exploring Melody Perception with TRACX2: A Recurrent Connectionist Model

Published on April 17, 2023

Imagine you’re in a music class and you’re learning to recognize different melodies. Just like how our brains chunk words and images, researchers have developed a clever model called TRACX2 to simulate how we perceive melodies. This model, which has been successfully used for speech and image processing, trains itself to recognize short sequences of intervals commonly found in French children’s songs. It then incorporates these musical chunks into new input, creating internal representations that resemble human-recognizable melodic categories. TRACX2 is sensitive to both the shape and proximity of these musical chunks. In fact, it even demonstrates an ‘end-of-word’ superiority effect similar to how we remember short musical phrases! These findings suggest that the same mechanism for chunking and segmentation applies not only to words and images, but also to elementary melody processing. If you’re interested in diving deeper into how TRACX2 works and its implications for music processing, check out the full article!

Abstract
Are similar, or even identical, mechanisms used in the computational modeling of speech segmentation, serial image processing, and music processing? We address this question by exploring how TRACX2, a recognition-based, recursive connectionist autoencoder model of chunking and sequence segmentation, which has successfully simulated speech and serial-image processing, might be applied to elementary melody perception. The model, a three-layer autoencoder that recognizes “chunks” of short sequences of intervals that have been frequently encountered on input, is trained on the tone intervals of melodically simple French children’s songs. It dynamically incorporates the internal representations of these chunks into new input. Its internal representations cluster in a manner that is consistent with “human-recognizable” melodic categories. TRACX2 is sensitive to both contour and proximity information in the musical chunks that it encounters in its input. It shows the “end-of-word” superiority effect demonstrated by Saffran et al. (1999) for short musical phrases. The overall findings suggest that the recursive autoassociative chunking mechanism, as implemented in TRACX2, may be a general segmentation and chunking mechanism, underlying not only word- and image-chunking, but also elementary melody processing.

Read Full Article (External Site)

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes:

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>