Upper Limb Motion Recognition Based On Bio Impedance Tomography

Adaptation to the operator's intention and external physical uncertainty I s the key factor that affects the performance of human-machine collaboration.In this paper,we propose a human-machine control interface that combines electrical impedance tomography(EIT)-based sensing methods with a robot controller to provide appropriate assistance to external uncertainties in collaborative tasks.The design and realization of a man-machine interface control system oriented to man-machine interaction.The interface first uses the optimized EIT features to estimate the continuous forearm muscle contraction(expressed in grasping force)for the wearable fabric belt,and fits the collected EIT data to the angle or grip strength data,verifying that the EIT man-machine interface is effective for the upper limbs.The performance indicators of the motion estimation method and the adaptability of the EIT signal characteristic algorithm.Secondly,the data acquisition method and the use of the host computer have the advantages of easy analysis and easy scientific research.Recognition and decision-making are then used as control input to adjust state transitions and desired interactions in real time.force.In the EIT experiment,the upper limb movement recognition based on the EIT signal,the continuous angle estimation experiment of the upper limb,the discrete gesture recognition experiment,the continuous grasping force estimation experiment,the continuous force estimation experiment of external disturbance,and the repeated wearing experiment were carried out to explore the electrical impedance.The tomography man-machine interface technology was finally applied to the UR5 robot platform.In grasping force estimation and man-machine sawing tasks,we recruited more than 20 subjects to evaluate the interface.For the grasping force estimation,the average r2 generated by the interface = 0.9,in offline and online verification and feature optimization procedures and an s-type regression function.The interface is robust to external interference without the need for retraining or manual calibration.The average r2 of the untrained dynamic posture is 0.86,and the average r2 value is between 0.85 and 0.88 in the test using the re-wearing front end.The performance of this interface can be comparable to or better than that of grasping force estimation based on bio-signals and man-machine sawing technology.Future efforts are worth paying in this new direction to achieve more promising results.