Carlos Eduardo Cancino Chacón

An Evaluation of Score Descriptors Combined with Non-linear Models of Expressive Dynamics in Music

Expressive interpretation forms an important but complex aspect of music, in particular in certain forms of classical music. Modeling the relation between musical expression and structural aspects of the score being performed, is an ongoing line of research. Prior work has shown that some simple numerical descriptors of the score (capturing dynamics annotations and pitch) are effective for predicting expressive dynamics in classical piano performances. Nevertheless, the features have only been tested in a very simple linear regression model. In this work, we explore the potential of a non-linear model for predicting expressive dynamics. Using a set of descriptors that capture different types of structure in the musical score, we compare the predictive accuracies of linear and non-linear models. We show that, in addition to being (slightly) more accurate, non-linear models can better describe certain interactions between numerical descriptors than linear models.

Probabilistic Segmentation of Musical Sequences Using Restricted Boltzmann Machines

A salient characteristic of human perception of music is that musical events are perceived as being grouped temporally into structural units such as phrases or motifs. Segmentation of musical sequences into structural units is a topic of ongoing research, both in cognitive psychology and music information retrieval. Computational models of music segmentation are typically based either on explicit knowledge of music theory or human perception, or on statistical and information-theoretic properties of musical data. The former, rule-based approach has been found to better account for (human annotated) segment boundaries in music than probabilistic approaches [14], although the statistical model proposed in [14] performs almost as well as state-of-the-art rule-based approaches. In this paper, we propose a new probabilistic segmentation method, based on Restricted Boltzmann Machines (RBM). By sampling, we determine a probability distribution over a subset of visible units in the model, conditioned on a configuration of the remaining visible units. We apply this approach to an n-gram representation of melodies, where the RBM generates the conditional probability of a note given its \(n-1\) predecessors. We use this quantity in combination with a threshold to determine the location of segment boundaries. A comparative evaluation shows that this model slightly improves segmentation performance over the model proposed in [14], and as such is closer to the state-of-the-art rule-based models.

Convolution-based classification of audio and symbolic representations of music

Abstract We present a novel convolution-based method for classification of audio and symbolic representations of music, which we apply to classification of music by style. Pieces of music are first sampled to pitch–time representations (spectrograms or piano-rolls) and then convolved with a Gaussian filter, before being classified by a support vector machine or by k-nearest neighbours in an ensemble of classifiers. On the well-studied task of discriminating between string quartet movements by Haydn and Mozart, we obtain accuracies that equal the state of the art on two data-sets. However, in multi-class composer identification, methods specialised for classifying symbolic representations of music are more effective. We also performed experiments on symbolic representations, synthetic audio and two different recordings of The Well-Tempered Clavier by J. S. Bach to study the method’s capacity to distinguish preludes from fugues. Our experimental results show that our approach performs similarly on symbolic representations, synthetic audio and audio recordings, setting our method apart from most previous studies that have been designed for use with either audio or symbolic data, but not both.