With the development of robotic technology, the demand for state-of-the-art technology in the field of rehabilitation is rapidly increasing for the elderly and people with disabilities. In this paper, we propose a real-time virtual coach to assist physical therapists with the end-effector-based robot-assisted gait training for stroke survivors using Long Short-Term Memory (LSTM) networks. Our proposed virtual coach consists of the sensor module for data gathering and dataset generation, real-time classification of the pathologic patient gait during the training using LSTM networks, and delivery of the coaching recommendations in an audiovisual form. Our preliminary study determined the selection of coaching recommendations. LSTM networks are trained to provide the selected coaching recommendations. The performance of the proposed virtual coach is verified using classification simulation of an able-bodied person on the rehabilitation robot, G-EO System. The usability was verified through a satisfaction survey of five professional physical therapists.

Robotic rollators (walkers with motor) require appropriate assist performance to improve the quality of life of the elderly and disabled. In this pilot study, the performance of commercial robotic rollators was evaluated by analyzing the effect of load condition on muscle activities and knee and hip joint angle when walking on a slope. Results showed that, the highest muscle activity occurred under a load of 5 kg but there was no difference in joint angle according to load condition. In future studies, we plan to test the performance with many subjects. In addition, we will perform motion analysis and muscle activity analysis according to changes in assist, brake, and speed levels.

A human who uses a SEMG-based power assist robot suffers from unstable assistance due to noisy nature of SEMG signals. This study aimed at analyzing the quantitative stability of the human elbow movements during the power assistance. During self-paced elbow flexion in the sagittal plane, an exoskeleton robot provided the assistive torque that was proportional to the estimated human elbow torque using SEMG. Maximum finite time Lyapunov exponent (MFTLE), the average logarithmic rate of the divergence of neighboring trajectories, during elbow flexion assisted by the robot was computed as an index of movement stability. The results showed a trade-off between decrease of physical effort for the movements and the stability of the movements. The stability of the SEMG-based assistance by MFTLE deteriorated as the amount of the assistive torque was increased to the amount of the human torque, although the physical effort required by the user his forearm was decreased. This study can be used as a guide to determine the amount of SEMG-based assistive torque for maintaining the stability of the assisted movements.

There has been a paucity of studies to identify the variation of muscle properties due to muscle fatigue, although feature changes of surface electromyographic signals due to fatigue have been reported on. In this paper, we investigated the variation of muscle properties in pre-fatigue condition and post-fatigue condition by using the SEMG and the dynamic muscle model. Five subjects performed index finger isometric abduction contraction by using the first dorsal interosseous (FDI) muscle. After fatigue-inducing contraction, the total twitch duration increased by 30.10%, the contraction time and half relaxation time (RT 1/2) were raised to 7.45% and 13.11%, respectively. These results indicate that the response of the twitch force was slowed and prolonged due to the fatigue-inducing contraction. Our results can be used to monitor and identify the muscle properties of patients in rehabilitation programs.

A current challenge with human-machine cooperation systems is to estimate human motions to facilitate natural cooperation and safety of the human. It is a logical approach to estimate the motions from their sources (skeletal muscles); thus, we employed surface electromyography (SEMG) to estimate body motions. In this paper, we investigated a cooperative manipulation control by an upper limb motion estimation method using SEMG and joint angular velocities. The SEMG signals from five upper limb muscles and angular velocities of the limb joints were used to approximate the flexion-extension of the limb in the 2-D sagittal plane. The experimental results showed that the proposed estimation method provides acceptable performance of the motion estimation [normalized root mean square error (NRMSE) <0.15, correlation coefficient (CC) >0.9] under the noncontact condition. From the analysis of the results, we found the necessity of the angular velocity input and estimation error feedback due to physical contact. Our results suggest that the estimation method can be useful for a natural human-machine cooperation control.

Due to difficulties in measurement of muscle activities and understanding a user's intention under different configurations, controlling machine forces using surface electromyogram (SEMG) is difficult in a human-machine interface (HMI). This study describes a novel HMI using Hill-based muscle model to control the isometric force of a robotic thumb that considers the importance of the thumb in hand function. In order to estimate force intension, SEMG from the skin surface was measured and converted to muscle activation information. The activations of deep muscles were inferred from the ratios of muscle activations from earlier studies. The muscle length of each contributed muscle was obtained by using a motion capture system and musculoskeletal modeling software packages. Once muscle forces were calculated, thumb-tip force was estimated based on a mapping model from the muscle force to thumb-tip force. The proposed method was evaluated in comparisons with a linear regression and artificial neural network (ANN) under four different thumb configurations to investigate the potential for estimations under conditions in which the thumb configuration changes.

Human motion and its intention sensing from noninvasive biosignals is one of the significant issues in the field of physical human-machine interactions (pHMI). This paper presents a real-time upper limb motion prediction method using surface electromyography (sEMG) signals for pHMI. The sEMG signals from 5 channels were collected and used to predict the motion by an artificial neural network (ANN) algorithm. We designed a human-machine interaction system to verify the proposed method. Interaction experiments were performed with or without physical contact, and the effects of instances of contact were investigated. The experimental results were compared with controlled experiments using a customized goniometer, which is able to measure upper limb flexion-extension. The results showed that the proposed method was not superior to the use of direct angle measurements; however, it provides sufficient accuracy and a fast response speed for interactions. SEMG-based interactions will become more natural with further studies of human-machine combination models.