Search

Recovering motor activation signals from peripheral nerve fascicles is the most acknowledged strategy for the control of prostheses driven by electroneurogram. Yet, restoring signals capturing correct muscular co-activation dynamics is still a challenge. This study aims to identify, through a novel hybrid paradigm, a reference model for muscular co-activation dynamics, whose parameters could be controlled by electroneurogram. A dataset related to an animal study was exploited. Muscular co-activation was modelled in the phase space using the Poincaré plot. The information of a selected electroneurogram channel about gait events was extracted and employed for the first time to design the Poincaré section. Utilising the intersection points, proposed iterated maps, X- and Y-maps, were trained to develop a reference model. The cross-validated prediction error of the model was 2.06 × 10-4 ± 0.0011 and 7.17 × 10-6 ± 4.59 × 10-5 for the X- and Y-maps, respectively. Analogous large- and small-scale patterns of the recurrence plots corresponding to sample model-generated and true data verified their similar dynamical regimes. Low variability of the model parameters, intra- and inter-trially, corroborated generalisability of the model. The viability of exploiting electroneurogram for modelling and predicting muscular co-activation was established. Leveraging reference model not only simplifies, but also opens up new possibilities for near-natural control of prostheses. [ABSTRACT FROM AUTHOR]