Task based motion intention prediction with EEG signals

This article attempts to review papers on power assist rehabilitation robots, human motion intention, control laws, and estimation of power assist rehabilitation robots based on human motion intention in recent years. This paper presents the various ways in which human motion intention in rehabilitation can be estimated. This paper also elaborates on the control laws for the estimation of motion intention of the power assist rehabilitation robot. From the review, it has been found that the motion intention estimation method includes: Artificial Intelligence-based motion intention and Model-based motion intention estimation. The controllers include hybrid force/position control, EMG control, and adaptive control. Furthermore, Artificial Intelligence based motion intention estimation can be subdivided into Electromyography (EMG), Surface Electromyography (SEMG), Extreme Learning Machine (ELM), and Electromyography-based Admittance Control (EAC). Also, Model-based motion intention estimation can be subdivided into Impedance and Admittance control interaction. Having reviewed several papers, EAC and ELM are proposed for efficient motion intention estimation under artificial-based motion intention. In future works, Impedance and Admittance control methods are suggested under model-based motion intention for efficient estimation of motion intention of power assist rehabilitation robot. In addition, hybrid force/position control and adaptive control are suggested for the selection of control laws. The findings of this review paper can be used for developing an efficient power assist rehabilitation robot with motion intention to aid people with lower or upper limb impairment.

A human motion intent estimation algorithm based on laser ranger finder and force sensors is designed for a walking-aid robot. By applying the proposed algorithm to the robot, the user can operate the walking-aid robot more fluently, comfortably and safely. A walking-aid or rehabilitation robot normally estimates the user's motion intention by measuring the interactive forces between human and robot. Whereas, the human motion should be represented as a coordination between the arms and the legs movements. This study therefore proposes a human motion intent estimation algorithm by fusing the motion intentions detected by the force sensors and the laser ranger finder, which are used to monitor the movements of arms and legs respectively. A special arrangement of push-pull force sensors is implemented for estimating the human motion intention I from user's operations on the handles. Meanwhile the human motion intention II is extracted by a laser ranger finder observing the user's movement of legs. A multi-sensor fusion algorithm is employed to fuse these two estimates. Experimental results are presented to show the validity of the proposed human motion intent estimation algorithm.

This paper presents the current state of research into power-assist exoskeletons for the upper limb. The assist of the upper limb is important for physically weak persons in daily activities, since upper-limb motion is involved in many important motions in daily living. The most important criterion is that power-assist exoskeletons assist the user's motion automatically in accordance with the user's motion intentions. Electromyogram (EMG) signals in which the user's motion intention is reflected could provide vital real-time information to facilitate accurate control of the power-assist exoskeleton in accordance with the user's motion intentions. A four degree-of-freedom active exoskeleton that assists human upper-limb motion (shoulder vertical flexion/extension, shoulder horizontal flexion/extension, elbow flexion/extension, and forearm supination/pronation) is also proposed.

It may not an easy task for physically weak elderly, disabled and injured individuals to perform the day to day activities in their life. Therefore, many assistive devices have been developed in order to improve the quality of life of those people in which they may not depend on others. Especially upper-limb power-assist exoskeletons have been developed since the upper limb motions are very important for the daily activities. Electromyography (EMG) signals and/or force sensor based control methods have been identified as the promising methods to control such exoskeleton devices. However, if the user cannot generate sufficient muscle signals or movements, the EMG or force sensor based methods could not be useful to the user. On the other hand, electroencephalography (EEG) signals are also important biological signals to extract the user's motion intention. In this study, the user's elbow joint motion is estimated based on the EEG signals. The measured EEG signals are pre-processed and input to a time-embedded linear model, which is assumed to decode the elbow joint angular velocities. The genetic algorithm (GA) is used to train the model. A six fold cross validation process was performed for each motion segment of each subject. The experimental results suggest that EEG signals with the tested decoding model can be used to continuously decode the elbow joint velocity.

Estimation of human impedance and motion intention for constrained human–robot interaction

In order to assist the motion in the daily lives of physically weak persons such as elderly persons, many kinds of power-assist robots have been developed. In the case of some physically weak persons, the ability to perceive the environment is sometimes deteriorated also. A method of perception-assist has been proposed to assist not only the user's motion but also the user's interaction with an environment, by applying the modification force to the user's motion if it is necessary. In this paper, the perception-assist for a lower-limb power-assist exoskeleton robot is proposed. In the daily life, the walking is very important for persons to achieve desired tasks. Basically, the robot assists the user's muscle force according to the user's motion intention which is estimated based on EMG signals. If the robot has found problems which might lead the user to dangerous situation such as the falling down, the robot tries to modify the user's motion in addition to the ordinal power-assists to make the user walk properly. Since the user might fall down by the effect of the additional modification force of the perception-assist, the robot automatically prevents the user from falling down by considering ZMP (Zero Moment Point). The effectiveness of the proposed method has been evaluated by performing experiments.

Major challenges of controlling human-robot collaboration (HRC)-oriented modular robot manipulators (MRMs) include the estimation of human motion intention while cooperating with a robot and performance optimization. This article proposes a cooperative game-based approximate optimal control method of MRMs for HRC tasks. A harmonic drive compliance model-based human motion intention estimation method is developed using robot position measurements only, which forms the basis of the MRM dynamic model. Based on the cooperative differential game strategy, the optimal control problem of HRC-oriented MRM systems is transformed into a cooperative game problem of multiple subsystems. By taking advantage of the adaptive dynamic programming (ADP) algorithm, a joint cost function identifier is developed via the critic neural networks, which is implemented for solving the parametric Hamilton-Jacobi-Bellman (HJB) equation and Pareto optimal solutions. The trajectory tracking error under the HRC task of the closed-loop MRM system is proved to be ultimately uniformly bounded (UUB) by the Lyapunov theory. Finally, experiment results are presented, which reveal the advantage of the proposed method.

In an ideal future limb function can be restored after suffering any kind of damage from an accident or a disease. The dream of fully functional prosthetic limbs depicts a great opportunity towards this vision. One of the main challenges is to create a control algorithm which processes physiological signals of the human body and moves the prosthetic limb according to the user's intentions. Brain-Computer Interfaces (BCIs) are popular subjects of research worldwide in the view of prosthesis control. Building mostly on EEG (electroencephalogram) signals, such systems must perform a challenging task - extracting control information from a signal with extremely low signal-to-noise ratio. Various digital signal processing and machine learning algorithms have been published in the past few years, but there is still no clear candidate for the title of the best one for e.g. detecting hand movement intention. In this paper various algorithms are compared on the example of detecting hand movement from EEG signals in real-time. The first goal of this paper is to present a couple of promising methods from other researchers in the past years, along with demonstrating the performance of other algorithms on a certain grasp-and-lift task. Another goal of the paper is to outline a future direction for the research of movement intent detection algorithms.

An electromyography signals-based human-robot collaboration system for human motion intention recognition and realization

A neural network-based model for lower limb continuous estimation against the disturbance of uncertainty

BackgroundFor severely paralyzed people, a brain-computer interface (BCI) provides a way of re-establishing communication. Although subjects with muscular dystrophy (MD) appear to be potential BCI users, the actual long-term effects of BCI use on brain activities in MD subjects have yet to be clarified. To investigate these effects, we followed BCI use by a chronic tetraplegic subject with MD over 5 months. The topographic changes in an electroencephalogram (EEG) after long-term use of the virtual reality (VR)-based BCI were also assessed. Our originally developed BCI system was used to classify an EEG recorded over the sensorimotor cortex in real time and estimate the user's motor intention (MI) in 3 different limb movements: feet, left hand, and right hand. An avatar in the internet-based VR was controlled in accordance with the results of the EEG classification by the BCI. The subject was trained to control his avatar via the BCI by strolling in the VR for 1 hour a day and then continued the same training twice a month at his home.ResultsAfter the training, the error rate of the EEG classification decreased from 40% to 28%. The subject successfully walked around in the VR using only his MI and chatted with other users through a voice-chat function embedded in the internet-based VR. With this improvement in BCI control, event-related desynchronization (ERD) following MI was significantly enhanced (p < 0.01) for feet MI (from -29% to -55%), left-hand MI (from -23% to -42%), and right-hand MI (from -22% to -51%).ConclusionsThese results show that our subject with severe MD was able to learn to control his EEG signal and communicate with other users through use of VR navigation and suggest that an internet-based VR has the potential to provide paralyzed people with the opportunity for easy communication.

Brain Computer Interfaces (BCIs) have several applications. One of the most crucial domains that BCIs cater to is prosthetics, orthotics and rehabilitation devices. Such devices aim to better the lives of specially-abled people. Motor Imagery or Movement Intention (MI) neural signals are collected to obtain a better insight of the underlying neural activities. EEG is the most popular modality for obtaining neural signals due to its advantages like non-invasiveness, portability, and the ease of use. The utilization of MI EEG signals to design BCIs has been heavily exploited. Spectral estimation methods have been exploited to delineate different MI tasks. This article focusses on exploring channel-wise temporal and spectral features of MI EEG signals. A rich MI EEG dataset involving five different motor imageries has been utilized in this study. The analysis based upon channel-wise temporal and spectral decomposition involves two major biomarkers, Event Related Spectral Perturbation (ERSP) and Inter Trial Coherence (ITC). ERSP can be considered as a generalization of power increase and decrease within a frequency range of interest. ITC reflects consistency of rhythms across the trials. This article aims to exploit ERSP and ITC methods to explore relevant time segment, frequency range and informative channels associated to MI signatures.

Human motion intention recognition is a key to achieve perfect human-machine coordination and wearing comfort of wearable robots. Surface electromyography (sEMG), as a bioelectrical signal, generates prior to the corresponding motion and reflects the human motion intention directly. Thus, a better human-machine interaction can be achieved by using sEMG based motion intention recognition. In this paper, we review and discuss the state of the art of the sEMG based motion intention recognition that is mainly used in detail. According to the method adopted, motion intention recognition is divided into two groups: sEMG-driven musculoskeletal (MS) model based motion intention recognition and machine learning (ML) model based motion intention recognition. The specific models and recognition effects of each study are analyzed and systematically compared. Finally, a discussion of the existing problems in the current studies, major advances, and future challenges is presented.

The objective of this research is to explore whether a two-dimensional BCI can be achieved by reliably decoding single-trial magneto-encephalography (MEG) signal associated with sustaining or ceasing right and left hand movements. Seven naïve subjects participated in the study. Signals were recorded from 275-channel MEG and synthetic aperture magnetometry (SAM) was employed. The multi-class classification for four-directional control was evaluated offline from 10-fold cross-validation using direct-decision tree classifier and genetic algorithm based Mahalanobis linear distance. Beta band (15-30Hz) event-related desynchronization and event related synchronization were observed in right and left hand movement related motor areas for physical movements as well as motor imagery. The cross-validation accuracy for the proposed four-direction classification from SAM- filtered MEG signal was as high as 95-97% for physical movements and 86-87% for motor imagery. The high classification accuracy suggests that a reliable high performance two-dimensional BCI can be achieved from single trial detection of human natural movement intentions from SAM-filtered MEG signals, where user may not need extensive training.

One challenge of designing robot-assisted control is to identify and estimate human intention force and motion under dynamic disturbances. We present a robotic control design to help human in manipulator-assisted upperlimb movement applications. The control design uses a muscle synergy-based neural network method to predict human force and intention motion. A disturbance observer-based controller is designed to eliminate the influence of disturbances and allows human operators to achieve task by using their normal efforts. The control design takes advantage of predicted human force and intention motion to provide proper assistance in the human-manipulator interactions. Human subject experiments are presented to demonstrate the control robustness and performance of disturbance rejection. The comparison results with a benchmark controller also confirm that the proposed design provides manipulator-assisted capability to save human effort when there are additional loads and unknown disturbances.