Compression-based Modelling of Musical Similarity Perception

Abstract

Similarity is an important concept in music cognition research since the similarity between (parts of) musical pieces determines perception of stylistic categories and structural relationships between parts of musical works. The purpose of the present research is to develop and test models of musical similarity perception inspired by a transformational approach which conceives of similarity between two perceptual objects in terms of the complexity of the cognitive operations required to transform the representation of the first object into that of the second, a process which has been formulated in informationtheoretic terms. Specifically, computational simulations are developed based on compression distance in which a probabilistic model is trained on one piece of music and then used to predict, or compress, the notes in a second piece. The more predictable the second piece according to the model, the more efficiently it can be encoded and the greater the similarity between the two pieces. The present research extends an existing information-theoretic model of auditory expectation (IDyOM) to compute compression distances varying in symmetry and normalisation using high-level symbolic features representing aspects of pitch and rhythmic structure. Comparing these compression distances with listeners' similarity ratings between pairs of melodies collected in three experiments demonstrates that the compression-based model provides a good fit to the data and allows the identification of representations, model parameters and compression-based metrics that best account for musical similarity perception. The compression-based model also shows comparable performance to the best-performing algorithms on the MIREX 2005 melodic similarity task.

Statistical learning and probabilistic prediction in music cognition: mechanisms of stylistic enculturation

“…Pearce and Müllensiefen tested this model by comparing compression distance with pairwise similarity ratings provided by listeners in three studies for stimuli consisting of one original pop melody and a manipulated version (containing rhythm, interval, contour, phrase order, and modulation errors) . The results showed very high correlations between compression distance and perceptual similarity (with coefficients ranging from 0.87 to 0.94), especially for IDyOM models configured to combine probabilistic predictions of pitch and timing.…”

“…56,113,114 Informally, IDyOM can be used to derive a compression distance D(x, y) between two musical stimuli x and y by training a model on x, using that model to predict y, and taking the average IC across all notes in y (see Ref. 115 for a formal presentation of the model). If x and y are very similar, the IC will be low; if they are very dissimilar, the IC will be high.…”

The MUST Set and Toolbox: A Set of 200 Musical Stimuli Varying in Balance, Contour, Symmetry, and Complexity: Behavioral and Computational Assessments

“…IDyOM has been shown to accurately predict Western listeners' pitch expectations in behavioral, physiological, and EEG studies (e.g., Egermann et al, 2013;Hansen & Pearce, 2014;Omigie, Pearce, & Stewart, 2012;Omigie, Pearce, Williamson, & Stewart, 2013;Pearce, 2005;Pearce, Ruiz, Kapasi, Wiggins, & Bhattacharya, 2010), even better than static rule-based models (e.g., Narmour, 1991;Schellenberg, 1997). It has also been proved to account for expectations of the timing of melodic events (Sauvé, Sayed, Dean, & Pearce, 2018) and harmonic movement (Sears, Pearce, Spitzer, Caplin, & McAdams, 2018;Harrison & Pearce, 2018), and to simulate other psychological processes in music perception, including similarity perception (Pearce & Müllensiefen, 2017), recognition Running head: THE MUST SET AND TOOLBOX memory (Agres, Abdallah, & Pearce, 2018), phrase boundary perception (Pearce, Müllensiefen, & Wiggins, 2010), and aspects of emotional experience (Egermann et al, 2013;Gingras et al, 2016;Sauvé et al, 2018). We used the IDyOM in two configurations: first, the short-term model (STM) that learns incrementally on each stimulus independently; second, adding to the STM a long-term model (LTM) trained on a large corpus of Western tonal music (903 folk songs and chorales; datasets 1, 2, and 9 from Table 4.1 in Pearce, 2005, comprising 50,867 notes): the BOTH configuration.…”

PPM-Decay: A Computational Model of Auditory Prediction with Memory Decay

“…PPM is a powerful sequence prediction algorithm that has proved well-suited to modeling the cognitive processing of auditory sequences (Agres et al, 2018;Barascud et al, 2016;Egermann et al, 2013;Pearce & Müllensiefen, 2017;Pearce et al, 2010;Pearce & Wiggins, 2006). In these contexts, PPM has traditionally been interpreted as an ideal observer, simulating an (approximately) optimal strategy for predicting upcoming auditory events on the basis of learned statistics.…”

“…The combination of the statistical power of PPM and the flexible decay kernel makes the model well-suited to simulating online auditory statistical learning under memory constraints and in changing statistical environments. A particularly relevant application domain is music cognition, which has already made significant use of PPM models without decay kernels (DiGiorgi, Dixon, Zanoni, & Sarti, 2017;Egermann et al, 2013;Harrison & Pearce, 2018;Pearce & Müllensiefen, 2017;Pearce et al, 2010;Pearce & Wiggins, 2006). Incorporating decay kernels into these models should be useful for capturing how recency effects and memory limitations influence the probabilistic processing of musical structure.…”