Fingers Working in Coordination: Hierarchy of EEG, EMG and Kinematics

Description

Humans use their fingers to mediate the majority of mechanical interactions between themselves and the world. They are complex mechanical systems, both in terms of system dynamics and corresponding control. The sophistication of their coordination is beyond what the most advanced robot is capable of. Decades of experimental studies have already identified and described individual components of fingers, laying the ground for full understanding of how they operate. However, the signal transmission from the central neural commands to the muscular system is a highly nonlinear dynamic multi-input, multi-output (MIMO) system. This hinders an integrative understanding of the control mechanism at the system level.The proposed study aims to develop a hierarchical model to elucidate the interaction between neural and muscular bases of finger kinematics. These bases, in the context of functional networks in electroencephalography (EEG) and spinal modules in electromyography (EMG), have been postulated as an intelligent approach employed by the brain to facilitate control of a redundant system. Yet, how these bases interact across different system levels is largely unclear. A hierarchical modeling approach, describing the general information transformation from EEG signals, spinal modules, to muscle activities related to finger movements in healthy subjects, may offer a more reliable model structure for the prediction of finger kinematics from EEG recordings, something potentially beneficial to mobile motor prosthetic applications.Another important objective of this proposed study is to investigate the changes in neural modules, muscle synergies, and nonlinear dynamics of neuromuscular system associated with motor learning. By applying state-of-art estimation techniques, we will identify these changes in a mathematical and systematic manner. Furthermore, we postulate that such learning is consistent with an optimization principle of the neural and muscular modules. We will develop an objective function based on these neural and muscular modules to provide a mechanistic understanding of such optimal motor learning. The success of our work will provide a generalized framework for designing control algorithm in future neural prostheses. The physiologically plausible models developed may likewise contribute to novel biologically inspired control method to enhance the dexterity of robots.