Nonlinear Identification and Time-Frequency Domain Analysis of Corticomuscular Responses During Card-Grabbing Using the NARMAX Method

The main objective of this paper is to identify the corticomuscular responses of card-grabbing and obtain a unified structural model. The unified structural model reveals the key characteristics shared across healthy subjects, quantifies human cortical sensorimotor function. Electroencephalogram (EEG) and electromyographic (EMG) signals of ten subjects was recorded during card-grabbing. Multiple forward regression orthogonal least squares (MFROLS) algorithm was used to detect a uniform model structure of the subjects, and a nonlinear autoregressive moving average model with exogenous inputs (NARMAX) corticomuscular model was established to analyze the human sensorimotor system in time-frequency domain. The corticomuscular model identified by MFROLS algorithm has both linear and nonlinear model terms, and achieved 0.9780 variance accounted for when implementing a one-step-ahead prediction. The peaking effect around 0, 12, 20 Hz was observed by frequency domain analysis. Using MFROLS algorithm can detect the common characteristics of corticomuscular responses during card-grabbing. The measured corticomuscular responses are the integration of sensorimotor feedback signals and motor commands in the cerebral cortex, and the human sensorimotor system has the characteristics of nonlinear harmonics at specific coupling frequencies.