Segmenting music through the joint estimation of keys, chords and structural boundaries

Abstract

In this paper, we introduce a new approach to music structure segmentation that is based on the joint estimation of structural segments, keys and chords in one probabilistic framework. More precisely, the boundaries of a structure segment are determined by detecting key changes and by utilizing the difference in prior probability of chord transitions according to their position in a structural segment. In contrast to many of the recent approaches to structural segmentation, this system does not work with self-similarity matrices, although it has been designed to integrate this kind of approach into the framework at a later stage. However, just the current version of the system, using only the estimated harmony, is already producing encouraging results, especially with respect to the precise localization of the boundaries.

A Literature Review on Music Parameter Extraction and Visualization

“…The EEG of music cognition has rarely been studied (ERPs); Koelsch et al (2000) determined that musical context, task relevance of accidental chords, degree of violation, and probability influenced music processing. Pauwels & Peeters (2013) provided a new approach to music structure segmentation based on an integrated estimate of structural segments, keys, and chords in a probabilistic framework. A priori probabilities of key changes and chord transitions define the boundaries of the structural segments.…”

Learning, Probability and Logic: Toward a Unified Approach for Content-Based Music Information Retrieval

“…Current PGM approaches in MIR dealing with multiple interrelated features generally resort to some manual "tricks" to make inference tractable. For instance, the model presented in Pauwels and Peeters (2013) jointly estimates chords with keys and structural boundaries using a HMM in which each state represents a combination of a key, a chord, and a structural position. The authors manually incorporate some musicologically motivated constraints in the transition matrix, which allows reducing the computation time of the Viterbi decoding step.…”

A feedback framework for improved chord recognition based on NMF-based approximate note transcription

“…Lee and Slaney [20] trained 24 key-specific HMMs corresponding to the 24 major/minor keys and selected the best model with high probability for a given audio signal. To deal with key changes, some studies tried to take into account the transition between adjacent keys [12,21,22]. Deep neural networks (DNNs) and recurrent neural networks (RNNs), which have significantly improved the accuracy of speech recognition, have recently been used for chord recognition [23,24].…”