Control of multiple DOFs robots using motor imagery EEG combined with Huffman coding

Abstract

A brain-computer interface (BCI) can help establish a bridge between humans and external devices, like assisting stroke patients with several daily tasks. Motor Imagery (MI) electroencephalogram plays a critical role in BCI applications. To address the problem of lack of control instruction sets for brain-computer interfaces based on motor imagery, this paper proposes the use of a variable-length coding strategy to control the movement of a KUKA robotic arm with multiple degrees of freedom (DOF). The motor imagery EEG signals are classified into four categories by using common spatial pattern (CSP) for feature extraction and support vector machine (SVM) for classification. The classification results of the left-hand and right-hand motor imagery are encoded with binary Huffman Coding, and the codewords are mapped to 7 degrees of freedom of the robotic arm. The number of occurrences of each Huffman codeword is used to control the rotation angle of the corresponding degree of freedom. The classification results of tongue and feet motor imagery are used to enter and exit the selection of the degrees of freedom of the robotic arm, respectively. In the designed object grasping and placing scene, the instruction control using Huffman variable-length coding strategy reduces the time by 19 and 18 times, and the efficiency increases by 23.4% and 20.2%, respectively, compared with the use of binary fixed-length coding. Experimental results demonstrate that the use of Huffman coding enables the control of multiple degrees of freedom of a robotic arm using finite motor imagery classification categories. The method can improve the information transfer efficiency and expand the set of control instructions for brain-computer interfaces, which provides new ideas for the application scenarios of motor imagery-based brain-computer interfaces.