Christophe Rhodes

Content-Based Music Information Retrieval: Current Directions and Future Challenges

The steep rise in music downloading over CD sales has created a major shift in the music industry away from physical media formats and towards online products and services. Music is one of the most popular types of online information and there are now hundreds of music streaming and download services operating on the World-Wide Web. Some of the music collections available are approaching the scale of ten million tracks and this has posed a major challenge for searching, retrieving, and organizing music content. Research efforts in music information retrieval have involved experts from music perception, cognition, musicology, engineering, and computer science engaged in truly interdisciplinary activity that has resulted in many proposed algorithmic and methodological solutions to music search using content-based methods. This paper outlines the problems of content-based music information retrieval and explores the state-of-the-art methods using audio cues (e.g., query by humming, audio fingerprinting, content-based music retrieval) and other cues (e.g., music notation and symbolic representation), and identifies some of the major challenges for the coming years.

Analysis of Minimum Distances in High-Dimensional Musical Spaces

We propose an automatic method for measuring content-based music similarity, enhancing the current generation of music search engines and recommended systems. Many previous approaches to track similarity require brute-force, pair-wise processing between all audio features in a database and therefore are not practical for large collections. However, in an Internet-connected world, where users have access to millions of musical tracks, efficiency is crucial. Our approach uses features extracted from unlabeled audio data and near-neigbor retrieval using a distance threshold, determined by analysis, to solve a range of retrieval tasks. The tasks require temporal features-analogous to the technique of shingling used for text retrieval. To measure similarity, we count pairs of audio shingles, between a query and target track, that are below a distance threshold. The distribution of between-shingle distances is different for each database; therefore, we present an analysis of the distribution of minimum distances between shingles and a method for estimating a distance threshold for optimal retrieval performance. The method is compatible with locality-sensitive hashing (LSH)-allowing implementation with retrieval times several orders of magnitude faster than those using exhaustive distance computations. We evaluate the performance of our proposed method on three contrasting music similarity tasks: retrieval of mis-attributed recordings (fingerprint), retrieval of the same work performed by different artists (cover songs), and retrieval of edited and sampled versions of a query track by remix artists (remixes). Our method achieves near-perfect performance in the first two tasks and 75% precision at 70% recall in the third task. Each task was performed on a test database comprising 4.5 million audio shingles.

Linked Data and You: Bringing Music Research Software into the Semantic Web

Abstract The promise of the Semantic Web is to democratize access to data, allowing anyone to make use of and contribute back to the global store of knowledge. Within the scope of the OMRAS2 Music Information Retrieval project, we have made use of and contributed to Semantic Web technologies for purposes ranging from the publication of music recording metadata to the online dissemination of results from audio analysis algorithms. In this paper, we assess the extent to which our tools and frameworks can assist in research and facilitate distributed work among audio and music researchers, and enumerate and motivate further steps to improve collaborative efforts in music informatics using the Semantic Web. To this end, we review some of the tools developed by the OMRAS2 project, examine the extent to which our work reflects the Semantic Web paradigm, and discuss some of the remaining work needed to fulfil the promise of online music informatics research.

Cosmological evolution of brane world moduli

We study cosmological consequences of non-constant brane world moduli in five dimensional brane world models with bulk scalars and two boundary branes. We focus on the case where the brane tension is an exponential function of the bulk scalar field, Ub∝exp(αϕ). In the limit α→0, the model reduces to the two-brane model of Randall-Sundrum, whereas larger values of α allow for a less warped bulk geometry. Using the moduli space approximation, we derive the four-dimensional low-energy effective action from a supergravity-inspired five-dimensional theory. For arbitrary values of α, the resulting theory has the form of a bi-scalar-tensor theory. We show that, in order to be consistent with local gravitational observations, α has to be small (less than 10−2) and the separation of the branes must be large. We study the cosmological evolution of the interbrane distance and the bulk scalar field for different matter contents on each branes. Our findings indicate that attractor solutions exist which drive the moduli fields towards values consistent with observations. The efficiency of the attractor mechanism crucially depends on the matter content on each branes. In the five-dimensional description, the attractors correspond to the motion of the negative tension brane towards a bulk singularity, which signals the eventual breakdown of the four-dimensional description and the necessity of a better understanding of the bulk singularity.

Branes on the Horizon

Models with extra dimensions are often invoked to resolve cosmological problems. We investigate the possibility of apparent acausality as seen by a brane-based observer resulting from signal propagation through the extra dimensions. Null geodesics are first computed in static and cosmological single-brane models, following which we derive the equations of motion for the inter-brane distance in a two-brane scenario, which we use to examine possible acausality in this more complex setup. Despite observing significant effective acausality in some situations there is no a priori solution to the horizon problem using this mechanism. In the two-brane scenario there can be significant late time violation of gravitational Lorentz invariance, resulting in the gravitational horizon being larger than the particle horizon, leading to potential signals in gravitational wave detectors.Models with extra dimensions are often invoked to resolve cosmological problems. We investigate the possibility of apparent acausality as seen by a brane-based observer resulting from signal propagation through the extra dimensions. Null geodesics are first computed in static and cosmological single-brane models, following which we derive the equations of motion for the inter-brane distance in a two-brane scenario, which we use to examine possible acausality in this more complex setup. Despite observing significant effective acausality in some situations there is no a priori solution to the horizon problem using this mechanism. In the two-brane scenario there can be significant late time violation of gravitational Lorentz invariance, resulting in the gravitational horizon being larger than the particle horizon, leading to potential signals in gravitational wave detectors.

Multiple Viewpoint Systems: Time Complexity and the Construction of Domains for Complex Musical Viewpoints in the Harmonisation Problem

Abstract We discuss the problem of automatic four-part harmonization: given a soprano part, add alto, tenor and bass in accordance with the compositional practices of a particular musical era. In particular, we focus on the development of representational and modelling techniques, within the framework of multiple viewpoint systems and Prediction by Partial Match (PPM), for the creation of statistical models of four-part harmony by machine learning. Our ultimate goal is to create better models, according to the information theoretic measure cross-entropy, than have yet been produced. We use multiple viewpoint because of their ability to represent both surface and underlying musical structure, and because they have already been successfully applied to melodic modelling. To allow for the complexities of harmony, however, the framework must be extended; for example, we begin by predicting complete chords, and then extend the framework to allow part by part prediction. As the framework is extended and generalized, the viewpoints become more complex. This article discusses matters related to viewpoint domains (alphabets), such as their size and consequent effect on run time; and presents methods for their reliable construction. We also present an empirical analysis of the time complexity of our computer implementation.

Using duration models to reduce fragmentation in audio segmentation

We investigate explicit segment duration models in addressing the problem of fragmentation in musical audio segmentation. The resulting probabilistic models are optimised using Markov Chain Monte Carlo methods; in particular, we introduce a modification to Wolff’s algorithm to make it applicable to a segment classification model with an arbitrary duration prior. We apply this to a collection of pop songs, and show experimentally that the generated segmentations suffer much less from fragmentation than those produced by segmentation algorithms based on clustering, and are closer to an expert listener’s annotations, as evaluated by two different performance measures.

Investigating music collections at different scales with AudioDB

Abstract Content-based search of music collections presents differing challenges at different scales and according to the task at hand. In this paper, we consider a number of different use cases for content-based similarity search, at scales ranging between a detailed investigation of a single track to searching for fragments of a track against a collection of millions of media items. We pay particular attention to the varying tradeoff between precision and recall in these contexts, both from the point of view of system evaluation and from the point of view of a user of a system searching an unknown collection. We present the audioDB software for content-based search, and describe how it has been used to address use cases across these different collection sizes; in addition we show that the interpretation of similarity as a distance which can be modelled statistically, initially motivated by our desire to achieve sublinear retrieval time on large databases, can be used to improve the precision of searches over small and medium-sized collections.

Bayesian Model Selection for Harmonic Labelling

We present a simple model based on Dirichlet distributions for pitch-class proportions within chords, motivated by the task of generating ‘lead sheets’ (sequences of chord labels) from symbolic musical data. Using this chord model, we demonstrate the use of Bayesian Model Selection to choose an appropriate span of musical time for labelling at all points in time throughout a song. We show how to infer parameters for our models from labelled ground-truth data, use these parameters to elicit details of the ground truth labelling procedure itself, and examine the performance of our system on a test corpus (giving 75% correct windowing decisions from optimal parameters). The performance characteristics of our system suggest that pitch class proportions alone do not capture all the information used in generating the ground-truth labels. We demonstrate that additional features can be seamlessly incorporated into our framework, and suggest particular features which would be likely to improve performance of our system for this task.

Analysis and Exploitation of Musician Social Networks for Recommendation and Discovery

This paper presents an extensive analysis of a sample of a social network of musicians. The network sample is first analyzed using standard complex network techniques to verify that it has similar properties to other web-derived complex networks. Content-based pairwise dissimilarity values between the musical data associated with the network sample are computed, and the relationship between those content-based distances and distances from network theory explored. Following this exploration, hybrid graphs and distance measures are constructed, and used to examine the community structure of the artist network. Finally, results of these investigations are shown to be mostly orthogonal between these distance spaces. These results are considered with a focus recommendation and discovery applications employing these hybrid measures as their basis.