A closed-loop brain–machine interface framework design for motor rehabilitation. (April 2020)

• Record Type: Journal Article
• Authors: Pan, Hongguang; Mi, Wenyu; Lei, Xinyu; Deng, Jun
• Is Part Of: Biomedical signal processing and control. Volume 58(2020)
• Publication Date: 2020-04
• Subjects: Brain–machine interface -- Framework design -- Auxiliary controller -- Network of spiking neurons -- Particle swarm optimization; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2020.101877
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A decoder based on the Wiener filter and an encoder based on a network of spiking neurons are designed to compensate for the absent information pathway, and a charge-balanced intra-cortical microstimulation current is chosen as the input of the spiking neuron network. Two auxiliary controllers are designed according to the strategy of model predictive control, where the controller inputs are the position of joint muscle trajectories and the average firing activity trajectories of perceived position vector neurons. Abstract: Brain–machine interfaces (BMIs) can be adopted to rehabilitate motor systems for disabled subjects by sensing cortical neuronal activities and creating new method. In this paper, to achieve the function of motor rehabilitation, two generalized BMI frameworks, including decoders, encoders and auxiliary controllers, are proposed and compared based on a classical single-joint information transmission model. Firstly, a decoder based on the Wiener filter and an encoder based on a network of spiking neurons are designed to compensate for the absent information pathway, and a charge-balanced intra-cortical microstimulation current is chosen as the input of the spiking neuron network; Secondly, to formulate closed-loop BMI frameworks, two auxiliary controllers are designed according to the strategy of model predictive control, where the controller inputs are the position of joint muscle trajectories and the average firing activity trajectories of perceived position vector neurons. Thirdly, considering that several integer parameters are included in the charge-balanced intra-cortical microstimulation current and that the optimization problem for solving the control inputs also includes these decision variables, a particle swarm optimization algorithm is adopted to solve the hard optimization problem. We compare the motor recovery effectiveness of the two presented frameworks through these simulations and choose the better framework for future BMI system design. The proposed frameworks provide a important theoretical guidance for designing BMI system applied in future life.