Biomechanics of Speech Motor Modules

自Bernstein（1967）的开创性工作以来，需要减少运动系统的维度，从而减少认知负荷，已被视为运动控制理论的核心挑战。越来越多的神经生理学工作提供的证据表明，身体运动的控制被分解为可管理的若干空间固定的神经肌肉模块（例如，Safavynia和Ting 2012）。最近在计算神经科学方面的工作提供了证据，证明神经系统使用这些模块来实现降维（Berger et al. 2013）。尽管最近有大量令人兴奋的运动系统模块化研究（包括精细运动系统的研究，如手指控制，例如Overduin et al. 2012），但与语音相关的领域基本上没有受到影响。这种遗漏可能部分是由于语音研究中的意识形态保守主义，但也与语音运动系统的肌肉结构（Sanders and Mu 2013）和控制空间（例如，Gick and Derrick 2009, Ghosh et al. 2010, Simko and Cummins 2010, Perkell 2012）相对较大的复杂性有关。模块化的绝大多数证据来自使用肌电图对神经肌肉活动的直接记录（见Kutch和Valero-Cuevas 2012）。然而，即使在相对不那么复杂的神经肌肉系统（Hug 2011, de Rugy et al. 2013）中，由于从肌电图中实验记录全面甚至具有代表性的神经肌肉活动的挑战（Hug 2011, de Rugy et al. 2013），我们预计生物力学将在获取语音产生基础的模块化神经肌肉结构方面发挥重要作用。在我们看来，神经肌肉模块是专门用来驱动生物力学上有效的身体结构的，使它们能够在很少或没有中央反馈控制的情况下进行前馈操作。这通常被认为是模块化的前提（例如，Loeb et al. 2000,2012），但很少经过测试（见Berniker et al. 2009的罕见例外），也从未应用于语音。语音生物力学建模的最新进展（例如，Stavness et al. 2012 a, b）使我们的团队能够开始识别一些我们认为是语音生成模块标志的生物力学特性（Gick et al. in press）。我们使用ArtiSynth （www.artisynth.org）进行了一系列声道模拟，实验结果支持了基于神经肌肉模块化的语音产生理论的可行性和实用性。这一建议将围绕认知科学中心问题的大量领域和方法联系起来，并对语音产生控制的许多方面，以及语音学，音韵学和语音发展（系统发育和个体发生）具有重大意义，其中周围神经结构和生物力学的细节迄今为止仍未被探索。神经肌肉水平的语言模块化将另一个运动系统带入神经模块的讨论，并有望在语音模型方面取得重大进展，在语言运动原语（Ramanarayanan et al. 2013）和新发现的与语音产生相关的皮层区域（Bouchard et al. 2013）之间构成潜在的神经肌肉“缺失环节”。

Since Bernstein’s (1967) groundbreaking work, the need to reduce the motor system’s dimensionality, and thereby to decrease cognitive load, has been seen as a central challenge for theories of motor control. A growing body of neurophysiology work has provided evidence suggesting that the control of body movements is broken down into a manageable number of spatially fixed neuromuscular modules (e.g., Safavynia and Ting 2012). Recent work in computational neuroscience provides evidence that the nervous system uses such modules to achieve dimensionality reduction (Berger et al. 2013). Despite a substantial and exciting recent body of research on modularization of motor systems (including work on fine motor systems, such as finger control, e.g., Overduin et al. 2012), speech-related fields have remained largely unaffected. This omission may be due partly to ideological conservatism in speech research, but it also relates to the relatively greater complexity of the muscular structures (Sanders and Mu 2013) and control space (e.g., Gick and Derrick 2009, Ghosh et al. 2010, Simko and Cummins 2010, Perkell 2012) of the speech motor system. The great majority of evidence for modularization derives from direct recordings of neuromuscular activity using electromyography (see Kutch and Valero-Cuevas 2012). However, because of the challenges of experimentally recording comprehensive or even representative neuromuscular activity from EMG, even in comparatively less complex neuromuscular systems (Hug 2011, de Rugy et al. 2013), we anticipate that biomechanics will play an important role in accessing the modular neuromuscular structures that underlie speech production. In our view, neuromuscular modules are built specifically to drive body structures that are biomechanically efficacious, enabling them to operate feed-forward with little or no central feedback control. This has often been assumed as a premise underlying modularization (e.g., Loeb et al. 2000, 2012), but has seldom been tested (see Berniker et al. 2009 for a rare exception), and never applied to speech. Recent advances in modeling speech biomechanics (e.g., Stavness et al. 2012 a, b) have enabled our group to begin identifying some of the biomechanical properties that we view as the hallmarks of speech production modules (Gick et al. in press). We present a series of vocal tract simulations using ArtiSynth (www.artisynth.org) and experimental results supporting the viability and utility of pursuing a theory of speech production built on neuromuscular modularization. This proposal links a large number of fields and methodologies surrounding a central question in cognitive science, and has dramatic implications for many aspects of speech production control, as well as for phonetics, phonology, and the development (phylogenetic and ontogenetic) of speech, where the details of peripheral neural structures and biomechanics have heretofore remained largely unexplored. Modularizing speech at the neuromuscular level brings another motor system into the discussion of neural modules, and promises a major advance for speech models, constituting a potential neuromuscular “missing link” between speech movement primitives (Ramanarayanan et al. 2013) and newly discovered cortical regions associated with speech production (Bouchard et al. 2013).