[论文解读] Modeling synchronization in human musical rhythms using Impulse Pattern Formulation (IPF)

Simon Linke, Rolf Bader+1 more|arXiv (Cornell University)|Dec 4, 2021

Music Technology and Sound Studies被引用 4

一句话总结

本文提出了脉冲模式公式（IPF），一种非线性递归模型，通过耦合脉冲列模拟节拍适应与波动，实现类人音乐节奏同步。IPF 成功复现了自然的同步动态，包括对节拍变化的响应和复合节拍耦合，优于固定节拍器，为现场与录音室音乐应用提供了逼真且自适应的替代方案。

ABSTRACT

In the publication, supplemented by these sound examples, the Impulse Pattern Formulation is used to model the synchronization of musicians to a collective tempo. Several click tracks are numerically created, representing eighth notes played by a musician or a metronome for different tempo changes.<br> By replacing every beat with a sound sample, audio files are created for a more musical evaluation of the results. The IPF is represented by a cowbell and the underlying click track with claves. Those sound files are in stereo, whereby the click track is at the left channel, and the IPF's signal is at the right channel. Thus, e.g., the balance potentiometer of a stereo system can be used to blend both sounds freely. The practical examples are: <strong>Fig.2:</strong><br> IPF reacts to four different step changes in tempo. <strong>Fig.5:</strong><br> The IPF reacting to the same changes in tempo as shown in Figure 2, when changing the tempo linear during 24 beats instead of step changes. <strong>Fig.8:</strong><br> IPF adapting to a noisy click track: the upper line (a) and b)) shows white noise, and the lower line (c) and d)) Brownian noise. On the left (a) and c)), the fluctuation is \(\pm1~\%\), and on the right (b) and d)) \(\pm 5~\%\). <strong>Fig.9:</strong><br> IPF adapting to a sinusoidally modulated click track: the upper line (a) and b)) shows a modulation period of 32 eighth notes, and the lower line (c) and d)) shows a modulation period of 8 eighth notes. On the left (a) and c)), the amplitude is 36 bpm, and on the right (b) and d)) 6 bpm, both centered around 113 bpm. <strong>Fig.12:</strong><br> Several scenarios shown in Figures 2, 5, 8, and 9 applied to an extended IPF which considers phase differences: a) step change from 120 to 100 bpm, b) linear change from 120 to 130 bpm, c) \(\pm 5~\%\) Brownian noise added to a 120 bpm click track and d) sinusoidal modulation with a period length of 32 eight notes varied \(\pm 6~bpm\) around 113 bpm. <strong>Fig.13:</strong><br> Several scenarios shown in Figures 2, 5, 8, and 9 applied to an extended IPF optimized for polyrhythms: a) step change from 120 to 140 bpm, b) linear change from 100 to 120 bpm, c) \(\pm 5~\%\) Brownian noise added to a 90 bpm click track and d) sinusoidal modulation with a period length of 32 eight notes varied \(\pm 6~bpm\) around 113 bpm. In all Figures, blue lines refer to the tempo of the click track, and red lines correspond to the tempo of the IPF. The single crosses represent single beats. A more in-depth description of how these sounds were synthesized can be found in the publication supplemented by these examples: Linke, S., Bader, R., &amp; Mores, R. (2021). Modeling synchronization in human musical rhythms using Impulse Pattern Formulation (IPF). http://arxiv.org/pdf/2112.03218v1

研究动机与目标

• 开发一种动态、自适应的固定节拍器与节拍器替代方案，更真实地反映音乐合奏中人类的自然同步行为。
• 利用受物理启发的数学框架，建模人类节奏同步的自组织非线性动力学。
• 以具备类人时间波动与自适应节拍追踪能力的系统，取代人工节拍器。
• 通过参数调节实现对个体节奏特征与创造性节奏行为的建模。
• 提供计算高效、支持实时运行的模型，适用于与现场乐手或数字音频工作站集成。

提出的方法

• IPF 使用非线性递归方程，建模代表节奏事件的指数衰减脉冲列之间的相互作用。
• 将同步建模为协同系统，其中演奏者与 IPF 系统的脉冲列通过反射点 βk 非线性耦合。
• 系统根据输入节奏动态调整输出节拍，模拟人类对节拍变化的自适应行为。
• 模型包含相变与瞬态响应，使其能够再现同步过程中的突变与滞后现象。
• 通过调节反射点 βk 和耦合强度等系统参数，可模拟不同节奏行为，包括复合节拍与个体时间特征。
• 通过将模型对受控节拍变化与复合节拍序列的响应，与既有的敲击实验数据及人类演奏数据对比，对模型进行验证。

实验结果

研究问题

• RQ1 IPF 模型能否准确模拟人类在音乐同步中对节拍变化的类人适应？
• RQ2 与固定节拍器相比，IPF 在复合节拍同步场景中的表现如何？
• RQ3 IPF 在多大程度上能复现人类音乐演奏中观察到的自然、自组织的时间波动？
• RQ4 IPF 是否可调节以模拟个体节奏特征，或生成新颖、富有表现力的节奏模式？
• RQ5 增加额外的反射点 βk 对模型的真实感与实时应用中的计算可行性有何影响？

主要发现

• IPF 在节拍发生阶跃变化时能迅速适应，表现出短暂的混沌瞬态，行为与人类演奏者相似。
• 在与复合节拍序列同步时，IPF 能够保持与底层节奏关系相对应的独立、离散的最终节拍比，证实其处理复杂节奏结构的能力。
• 在标准设置下，模型的有效节拍范围为 108–135 bpm，当多个复合节拍比同时激活时，带宽有所减小。
• 增加反射点 βk 显著提升了模型捕捉细微节奏动态的能力，但复杂度提升并不总能改善实时性能。
• IPF 在模拟人类同步方面优于固定节拍器，因其能响应节拍变化而非强制施加固定节拍。
• 该模型可调节以模拟真实的人类时间感，或生成新颖、富有表现力的节奏行为，展现出超越传统节拍器的创作潜力。