Prior research suggests listeners' autonomic variables may entrain with musical phrase arcs. Here, we apply a computational technique for automatic extraction of probabilistic phrase arc boundaries from recorded music audio and compare them to those derived from physiological signals (respiration and RR intervals). The objective is to automate the evaluation of synchronisation of autonomic reactions to musical phrases, thereby assessing music's ability to regulate physiological states.
20 participants' (10 women, aged 39±11 years old) physiological signals (respiration and RR intervals) were recorded whilst listening to Prokofiev playing Prokofiev's Gavotte Op.12 No.2 rendered on a reproducing piano. Loudness, in sones, is calculated from the recorded music audio, and smoothed and envelope values are derived from it. Prior inspection shows the musical phrases to have mean length 10s, [min=7.2s, max=17.5s]. The beats are automatically extracted after audio-score alignment. The physiological data is interpolated to match the music beats. Phrase arc boundary credence, with arcs assumed to be quadratic, are computed from the loudness profile and interpolated physiological signals using a novel Bayesian approach incorporating dynamic programming. The boundary credence profiles (Cr) are cross-correlated with optimal lag.
Correlation with loudness Cr returned mean Pearson's coefficient and standard deviation of 0.15±0.05 for respiration Cr's and 0.16±0.06 for RR Cr's. Higher correlation values occurred for positive offsets of [1,5] beats, reflecting a delay in physiological response to loudness. The maximum R for loudness and RR Cr's was 0.32, whilst for loudness and respiration Cr's, was 0.23, suggesting potentially higher degrees of synchronisation between the cardiovascular system and loudness compared to the respiratory system.
We have presented a fully automated data-driven technique for testing hypotheses about entrainment between musical phrase arcs and autonomic variables. Preliminary findings support the potential of employing music with specific structural qualities to modulate physiological signals.