Data Driven Symbiotic Machine Learning for the Identification of Motion-Based Action Potentials

Abstract

Understanding and modelling technical and biological processes is one of the basic prerequisites for the management and control of such processes. With the help of identification, the interdependencies of such processes can be deciphered and thus a model can be achieved. The verification of the models enables the quality of the models to be assessed. This article focuses on the identification and verification of motion and sensory feedback-based action potentials in peripheral nerves. Based on the acquisition of action potentials, the identification process correlates physiological and motion-based parameters to match movement trajectories and the corresponding action potentials. After a brief description of a prototype of a biosignal acquisition and identification system, this article introduces a new identification method, the symbiotic cycle, based on the well-known term symbiotic simulation. As an example, this article presents a data-driven method to create a human readable model without using presampled data. The closed-loop identification method is integrated into this symbiotic cycle.