7 February 2022 in IEEE Robotics and Automation Letters
Exoskeleton robots need to always actively assist the user’s movements otherwise robot just becomes a heavy load for the user. However, estimating diversified movement intentions in a user’s daily life is not easy and no algorithm so far has achieved that level of estimation. In this study, we rather focus on estimating and assisting a limited number of selected movements by using an EMG-based movement classification and a newly developed lightweight exoskeleton robot. Our lightweight knee exoskeleton is composed of a carbon fiber frame and highly backdrivable joint driven by a pneumatic artificial muscle. Thus, our robot does not interfere with the user’s motions even when the actuator is not activated. As the classification method, we adopted a positive-unlabeled (PU) classifier. Since precisely labeling all the selected data from large-scale daily movements is not practical, we assumed that only part of the selected data was labeled and used a PU classifier that can handle the unlabeled data. To validate our approach, we conducted experiments with five healthy subjects to selectively assist sit-to-stand movements from four possible daily motions. We compared our approach with two classification methods that assume fully labeled data. The results showed that all subject’s movements were properly assisted.
4 November 2019 in Journal of NeuroEngineering and Rehabilitation
Background: Muscle synergies are now widely discussed as a method for evaluating the existence of redundant neural networks that can be activated to enhance stroke rehabilitation. However, this approach was initially conceived to study muscle coordination during learned motions in healthy individuals. After brain damage, there are several neural adaptations that contribute to the recovery of motor strength, with muscle coordination being one of them. In this study, a model is proposed that assesses motion based on surface electromyography (sEMG) according to two main factors closely related to the neural adaptations underlying motor recovery: (1) the correct coordination of the muscles involved in a particular motion and (2) the ability to tune the effective strength of each muscle through muscle fiber contractions. These two factors are hypothesized to be affected differently by brain damage. Therefore, their independent evaluation will play an important role in understanding the origin of stroke-related motor impairments. Results: The model proposed was validated by analyzing sEMG data from 18 stroke patients with different paralysis levels and 30 healthy subjects. While the factors necessary to describe motion were stable across heathy subjects, there was an increasing disassociation for stroke patients with severe motor impairment. Conclusions: The clear dissociation between the coordination of muscles and the tuning of their strength demonstrates the importance of evaluating these factors in order to choose appropriate rehabilitation therapies. The model described in this research provides an efficient approach to promptly evaluate these factors through the use of two intuitive indexes.
17 July 2018 in Frontiers in Neurorobotics
An important function missing from current robotic systems is a human-like method for creating behavior from symbolized information. This function could be used to assess the extent to which robotic behavior is human-like because it distinguishes human motion from that of human-made machines created using currently available techniques. The purpose of this research is to clarify the mechanisms that generate automatic motor commands to achieve symbolized behavior. We design a controller with a learning method called tacit learning, which considers system–environment interactions, and a transfer method called mechanical resonance mode, which transfers the control signals into a mechanical resonance mode space (MRM-space). We conduct simulations and experiments that involve standing balance control against disturbances with a two-degree-of-freedom inverted pendulum and bipedal walking control with humanoid robots. In the simulations and experiments on standing balance control, the pendulum can become upright after a disturbance by adjusting a few signals in MRM-space with tacit learning. In the simulations and experiments on bipedal walking control, the robots realize a wide variety of walking by manually adjusting a few signals in MRM-space. The results show that transferring the signals to an appropriate control space is the key process for reducing the complexity of the signals from the environment and achieving diverse behavior.
21 February 2020 in IEEE Robotics and Automation Letters
Averaging electromyographic activity prior to muscle synergy computation is a common method employed to compensate for the inter-repetition variability usually associated with this kind of physiological recording. Capturing muscle synergies requires the preservation of accurate temporal and spatial information for muscle activity. The natural variation in electromyography data across consecutive repetitions of the same task raises several related challenges that make averaging a non-trivial process. Duration and triggering times of muscle activity generally vary across different repetitions of the same task. Therefore, it is necessary to define a robust methodology to segment and average muscle activity that deals with these issues. Emerging from this need, the present work proposes a standard protocol for segmenting and averaging muscle activations from periodic motions in a way that accurately preserves the temporal and spatial information contained in the original data and enables the isolation of a single averaged motion period. This protocol has been validated with muscle activity data recorded from 15 participants performing elbow flexion/extension motions, a series of actions driven by well-established muscle synergies. Using the averaged data, muscle synergies were computed, permitting their behavior to be compared with previous results related to the evaluated task. The comparison between the method proposed and a widely used methodology based on motion flags, shown the benefits of our system maintaining the consistency of muscle activation timings and synergies.
6 April 2022 in The Anatomical Record
Estimation of muscle activity using surface electromyography (sEMG) is an important non‐invasive method that can lead to a deeper understanding of motor‐control strategies in humans. Measurement using multiple active electrodes is necessary to estimate not only surface muscle activity but also deep muscle activity in dynamic motion. In this paper, we propose a method for estimating muscle activity of dynamic motions based on anatomical knowledge of muscle structures. To estimate muscle activity, a large number of signal sources are set in the muscle model, and connections between the signal sources are defined a priori based on the anatomical structure of the muscles. The signal source activities are first estimated by minimizing the Kullback–Leibler divergence with a continuity cost. Then, the muscle activity is computed from the signal source activity. In the experiments, five healthy participants performed five types of motion and the forearm sEMG was measured with 20‐channel active electrodes. The estimation results for these motions were visualized in four dimensions as the three‐dimensional position of the muscle over time. The results showed that the estimation was accurate, with a reproduction rate of 95% for the measured sEMG and continuity of the muscle activity. In addition, the results suggest the advantage of the proposed method over the conventional approaches in terms of estimating the muscle activity for both dynamic and abnormal motions.
26 June 2019 in Advanced Robotics
Robot rehabilitation is now recognized as an important method for the efficient recovery. In European Project FP7 BioMot, we have discussed the potential of the robot rehabilitation and proposed the suitable process for it. In this paper, we describe the proposed rehabilitation process and create the theoretical basis for the robot rehabilitation through designing the control system and the patient model. To design the patient model, we describe the source of paralysis and motion controller separately and define the recovery function from the paralysis. In the theoretical analysis of the control system, we show that the robot motions are first adapted to the patient abnormal motions and gradually drive the patient motions to the better ones by the motion support. The singular perturbation analysis proves that the stabilities of the two different process, adaptation to the patient motions and the motion support to the better ones, as a slow motion subsystem and a fast motion subsystem, respectively. The simulation results show that the proposed control system can drive the patients to the better state depending on the patient conditions such as recovery speed and recovery potential. The proposed system can be tuned to fit to the variety of the real patient conditions when we apply it to the real applications. GRAPHICAL ABSTRACT