Nonlinear audio recurrence analysis with application to genre classification

Abstract

In this paper we apply nonlinear signal analysis to a music information retrieval task. More concretely, we apply the concept of recurrence plots and recurrence histograms to extract information from music audio frames. We evaluate the effectiveness of this approach with a typical genre classification framework and compare it against a baseline obtained from standard spectrum-based descriptors. The accuracy reached by the histogram-based descriptors alone does not surpass the one achieved by the spectral-based descriptors. However, we show that the combination of both descriptor sources results in consistent improvements up to 5 absolute percent points. This highlights the potential of nonlinear signal analysis for quantitative music description. In particular, it suggests that the information resulting from this approach is complementary to the information obtained through the commonly used spectral representation.

The State of the Art Ten Years After a State of the Art: Future Research in Music Information Retrieval

“…Hartmann [28] notes finding seven duplicates, but mentions no specifics. In [15,56,71,74], the authors describe GTZAN as having 993 or 999 excerpts. In personal communication, de los Santos mentions that they found seven corrupted files in Classical (though [71,74] report them being in Reggae).…”

“…In [15,56,71,74], the authors describe GTZAN as having 993 or 999 excerpts. In personal communication, de los Santos mentions that they found seven corrupted files in Classical (though [71,74] report them being in Reggae). Li and Chan [46], who manually estimate the key of all GTZAN excerpts, mention in personal communication that they remember hearing some repetitions.…”

“…For Classify, most works measure MGR performance by classification accuracy (the ratio of "correct" predictions to all observations) computed from k-fold stratified cross-validation (kfCV), e.g., 2fCV (4 papers) [7,22,23,56], 3fCV (3 papers) [18,71,74], 5fCV (6 papers) [3,13,30,31,53,100], and 10fCV (55 papers) [2,5,9,11,14,16,17,24-26,28,29,34,35,37,39-42, 44,47-51,57,58,60-64,66-68,70,72,73,75,76,78,79,82-85,88-91,94-96,98,99]. Most of these use a single run of cross-validation; however, some perform multiple runs, e.g., 10 independent runs of 2fCV (10x2CV) [56] or 20x2fCV [22,23], 10x3fCV [71,74], and 10x10fCV [37,70,72,75,[83][84][85]. In one experiment, Li and Sleep [42] use 10fCV with random partitions; but in another, they partition the excerpts into folds based on their file number -roughly implementing an artist filter.…”

Improvisation and the self-organization of multiple musical bodies

“…Dynamical models have been employed in understanding how musicians improvise (Pressing, 1999, 2000) and coordinate with rhythmic sequences (Loehr et al, 2011), but David Borgo’s 2005 book Sync or Swarm is the most detailed account of exploring the skill of musical improvisation using the statistical tools of complex dynamical systems theory. More recent applications of these methods to examine musical movements and musical structure include: fractal analysis (Beauvois, 2007; Rankin et al, 2009; Demos et al, 2014; Hennig, 2014; Ruiz et al, 2014) recurrence quantification analysis (Serrà et al, 2009, 2011; Demos et al, 2011) and sample or Shannon entropy (Keller et al, 2011; Goldman, 2013; Glowinski et al, 2013). …”

Musical Collaboration in Rhythmic Improvisation

“…For example, Ernîo Lendvai identified the presence of Fibonacci numbers and golden ratios in many of Béla Bartók’s pieces [ 4 ]. Musical structures can be visualized and quantified by studying self-similarity over time from recurrent patterns [ 5 , 6 , 7 ] or constructing networks on the basis of the pitch and duration of notes [ 8 ]. Interestingly, recurrence-quantification analysis revealed that compositions of Bach’s inventions and sinfonias are more complex than a mere Markov process [ 9 ].…”