The analysis of electromyographic (EMG) signals is crucial for advancing human-machine interaction. Despite recent progress, most methods still approach gesture intention prediction as a single classification task, which overlooks the complex temporal dynamics and channel-specific variations present in EMG signals. To address these shortcomings, we propose FESNet (Fine-grained Electromyography Segmentation Network), a novel segmentation-based network that temporally segments EMG signals, enabling a more fine-grained analysis of finger movements. Our approach utilizes a robust backbone network for feature extraction, followed by a functional head that adapts to different granularities or task objectives (classification or segmentation). We evaluate our method on the Ninapro DB8 dataset, where FESNet outperforms previous models, demonstrating its superior performance. The source code is publicly available at: https://github.com/Dianli97/FESNet

Efficient processing of multichannel biosignals has significant application values in the fields of healthcare and human-machine interaction. Although previous research has achieved high recognition performance with deep convolutional neural networks, several key challenges still remain: (1) Effective extraction of spatial and temporal features from the multichannel biosignals. (2) Appropriate trade-off between performance and complexity for improving applicability in real-life situations given that traditional machine learning and 2D-based CNN approaches often involve excessive preprocessing steps or model parameters; and (3) Generalization ability of neural networks to compensate for domain difference and to reduce overfitting during training process. To address challenges 1 and 2, we propose a 1D-based deep intra and inter channel (I2C) convolution neural network. The I2C convolutional block is introduced to replace the standard convolutional layer, further extending it to several state-of-the-art modules, with the intent of extracting more effective features from multichannel biosignals with fewer parameters. To address challenge 3, we integrate a branch model into the main model to perform dynamic label smoothing, enabling the model to learn domain difference and improve its generalization ability. Experiments were conducted on three public multichannel biosignals databases, namely ISRUC-S3, HEF and Ninapro-DB1. The results suggest that the proposed method exhibits significant competitive advantages in accuracy, complexity, and generalization ability.

Biological signals classification is essential for human machine interaction. Although previous research has achieved high classification performance, compensating for domain shift due to the intra and inter individual variations remains a challenge. In this paper, we propose a novel dynamic label smoothing strategy, named DLS, to address this issue. The proposed DLS constructs an auxiliary neural network to adjust the true label and to supervise the primary neural network. Experiments on the NinaPro DB1 dataset demonstrate that the proposed DLS outperforms current state-of-the-art methods. Furthermore, the proposed DLS has significant potential for practical applications as it can maintain or even improve the performance of the primary neural network on noisy data. The source code is publicly available at: https://github.com/peijii/DLS

In this study, we propose a design methodology for anthropomimetic robotic wrists and forearms. Conventional robotic wrists and forearms have few examples of human mimicry, and they have been simplified. The ligaments and tendons of the robotic forearm proposed in this study were replicated using chain-stitched wires and arranged in a manner similar to the human anatomy. Motion capture and goniometer measurements were used to measure the range of motion (ROM) of the wrist and forearm, driven by 16 servomotors. The results of the experiment were compared with the human ROM reported in previous studies, and it was found that the developed robotic forearm could achieve a ROM similar to that of humans. The use of this robotic forearm in functional verification experiments will enhance our understanding of the human structure. The integration of mechanical mechanisms with structures created through evolution is expected to improve the functionality of future robots.

The activities of muscles in the forearm have been widely investigated to develop human interfaces involving hand motions, especially in the fields of prosthetic hands and teleoperation. Although surface electromyography (sEMG) is considered as an effective biological signal from which hand motions can be recognized, the availability and quality of sEMG data can limit the usability and intuitiveness of human interfaces. This article introduces force myography (FMG) as a supplementary signal and proposes a layered sEMG–FMG hybrid sensor that can measure both sEMG and FMG at the same skin surface location. Meanwhile, a layer fusion convolution neural network (LFC) is designed to extract multiscale features from sEMG and FMG. To evaluate the effectiveness of the hybrid sEMG–FMG sensor and LFC, a 22-hand motion classification experiment was conducted on nine able-bodied subjects. The recognition results indicated a significantly improved classification accuracy (p < 0.001) of the hybrid sEMG–FMG modality with respect to single sEMG or FMG modality. The classification accuracies (CAs) of LFC were compared with conventional machine learning methods, including support vector machine, random forest classifier, xgboost, and k-nearest neighbor. Compared with the single-modality sEMG, the CAs of the dual-modality sEMG–FMG using conventional methods, and LFC were improved by 21.31% and 16.71%, respectively. These results suggest that the layered sEMG–FMG sensing approach can effectively enhance the performance of human interfaces, which offers great potential in the clinical applications of sophisticated prosthetic hands and teleoperation.

This paper proposes a coupled tendon-driven waist joint for humanoid robots. The waist joint was designed as a 3 degrees of freedom (DOF) structure to simulate the motion of a human waist. The power transmission was designed by adopting a 3-motor 3-DOF (3M3D) coupled tendon-driven mechanism, so that the torque on the joints was multiplied. We derived the torque transmission formula and the rotation angle formula of the 3M3D tendon-driven structures and designed the waist joint by adopting an appropriate structure according to their features. To evaluate the accuracy and load capacity of the waist joint, we performed a rotational accuracy experiment and a maximum torque experiment. The experiment results showed that the maximum error of joint rotation was below 1°, and the maximum torque of the pitch, roll, and yaw rotations were 87[Nm], 53[Nm], and 22.2[Nm], respectively.

This article proposes coupled tendon-driven joint modules for anthropomorphic robots. Fully actuated 2-degree-of-freedom (DoF) and 3-DoF joint modules are classified and analyzed based on the motor-joint routing matrix that describes the tendon routing structure between the motors and the joints. Two-motor 2-DoF (2M2D) modules, which share the same form of motor-joint routing matrix, are classified into four types: externally actuated, internally coaxially actuated, internally separately actuated, and hybrid-actuated according to the location of the motors. Three-motor 3-DoF (3M3D) modules are classified into four forms based on the four possible forms of motor-joint routing matrix: fully routed motor-joint, 1-unrouted motor-joint, 2-unrouted motor-joint, and 3-unrouted motor-joint. The 2M2D and 3M3D coupled tendon-driven joint modules are analyzed and compared with respect to the relationship between the motor torque and the joint torque. A 7-DoF anthropomorphic robot arm was implemented with one 3M3D module for the shoulder joint and two 2M2D modules for the elbow and wrist joints, respectively, to demonstrate the utility of the proposed joint modules. The arm weighed 2.2 kg and was able to lift a 1.5-kg load with full outreach. The characteristics of the joint modules were evaluated in a current consumption experiment and a position accuracy experiment, and the performance of the robot arm was evaluated in a master-slave manipulation experiment involving dexterous movements.

This paper presents a gear-driven prosthetic hand designed for toddlers with transradial amputation. The hand design considers three main issues: weight, cost, and operability. The prosthetic hand and the cosmetic silicon glove are made based on the dimensions of a real hand. The simple, stable, and reliable gear-driven transmission helps to reduce the weight and the cost. A small actuator is embedded in the palm. During the grasp, the four fingers and the thumb flexes and extends as a unit to provide a wide range of holding area. The kinematics and static analysis in grasping was performed and the simulation results were compared with measured data. The motion performance and practical operability of the proposed hand was verified experimentally by a test system and a transradial subject.

This paper presents a lightweight (<;250 g) and low-cost (<;350 USD) biomimetic prosthetic hand with two actuators embedded in the palm. One of them is employed for flexion/extension of the five digits, and the other one is used for the adduction/abduction of thumb. Thus, the hand can achieve major grasping tasks that account for about 85% of activities in daily life. The unique transmission provides various advantages such as a compact structure, weight saving, and short driven distance. Furthermore, by using 3D printing technology, most parts of the prosthetic hand were made to be much lighter and have a humanlike appearance, compared with conventionally manufactured artificial hand. Finally, the performance and practical applicability of the proposed design was verified experimentally through both of a motion verification and an intuitive control test by a healthy subject and a transradial amputee.

A 2 motor 2 degrees-of-freedom (2M2D) coupled tendon driven joint module is proposed as a basic component for robot arms. Torque reallocation via tendon coupling can enhance the output torque of one single joint. According to the motor position, the joint module is classified into four types: the externally-actuated structure, the internally-coaxially-actuated structure, the internally-separately-actuated structure, and the hybrid-actuated structure. The four structures are analyzed and compared, and their implementation design examples are given. Experiments comparing the proposed joint module with directly-actuated traditional joint suggested that the 2M2D coupled tendon-driven joint module can obtain high control accuracy, and the torque reallocation via tendon coupling is effective to improve output torque. Additionally, an anthropomorphic robot arm with low weight and high payload was developed to show the utility of the proposed joint module.

Implantable brain machine interfaces (BMI) enable severely disabled people high-performance real-time robot control and communication, utilizing high-quality intracranial neural signals. Electrocorticograms (ECoG) are useful for implantable BMIs because of not only their zero time-lag property but their high spatiotemporal resolution with long term stability also. Fully implantable devices for ECoG recording offer long-term home-use with 24/7 supports. This will help not only patients with restoring motor and communication control but also help their caregivers with reducing burdens of caregiving day and night. Until now, we established ECoG-based robot control and communication. High gamma activity (80-150 Hz) was a good decoding feature for ECoG-based real time decoding and control. Independent component analyses effectively extract neural information with dimensional reduction and contribute to improving decoding accuracy. Also, we are developing a 128-channel fully-implantable BMI device (WHERBS) for long-term home-use with 24/7 supports. We completed GLP tests and non-clinical long-term implantation. The next step is a clinical trial to confirm safety and efficacy of the implantable BMI.

The dexterity of body motion when performing skills are being actively studied. In this paper, singular value decomposition is used to extract the dexterous features from the time-series data of body motion. A matrix is composed by overlapping the subsets of the time-series data. The left singular vectors of the matrix are extracted as the patterns of the motion and the singular values as a scalar, by which each corresponding left singular vector affects the matrix. A gesture recognition experiment, in which we categorize gesture motions with indexes of similarity and estimation that use left singular vectors, was conducted to validate the method. Furthermore, in order to understand the features better, the features of the left singular vectors were described as fuzzy sets, and fuzzy if-then rules were used to represent the knowledge.

The knowledge remembered by the human body and reflected by the dexterity of body motion is called embodied knowledge. In this paper, we propose a new method using singular value decomposition for extracting embodied knowledge from the time-series data of the motion. We compose a matrix from the time-series data and use the left singular vectors of the matrix as the patterns of the motion and the singular values as a scalar, by which each corresponding left singular vector affects the matrix. Two experiments were conducted to validate the method. One is a gesture recognition experiment in which we categorize gesture motions by two kinds of models with indexes of similarity and estimation that use left singular vectors. The proposed method obtained a higher correct categorization ratio than principal component analysis (PCA) and correlation efficiency (CE). The other is an ambulation evaluation experiment in which we distinguished the levels of walking disability. The first singular values derived from the walking acceleration were suggested to be a reliable criterion to evaluate walking disability. Finally we discuss the characteristic and significance of the embodied knowledge extraction using the singular value decomposition proposed in this paper.

Motion time-series data observed with various sensing systems are usually analyzed to extract embodied knowledge which is remembered by the human body and reflected by the dexterity in the motion of the body. A method based on singular value analysis through divided time-series data (SVA-DTS) is proposed for extracting features from time-series data. Matrices are composed from the subsets of time-series data and the left singular vectors of the matrices are extracted as the patterns of the motion and the singular values as a scalar, by which the corresponding left singular vectors affects the matrices. The SVA-DTS was applied to a walking difficulty evaluation experiment in which three levels of walking difficulty were simulated by restricting the right knee joint. The accelerations of the middles of the shanks and the back of the waist were measured. Singular values were calculated from the normalized acceleration time-series data with the SVA-DTS. The results showed that the first singular values inferred from the acceleration data of the right shank significantly related to the increase of the restriction to the right knee. The first singular values of the acceleration data of the right shank were suggested to be reliable criteria to evaluate walking difficulty. We visualize the first singular values in a 3D space to provide intuitive information about walking difficulty which can be used as a tool for evaluating walking difficulty.

Embodied knowledge is the knowledge remembered by the human body and reflected by the dexterity in the motion of the body. In this paper, we propose a new method using singular value decomposition for extracting embodied knowledge from the time-series data of the motion which is measured with various sensors such as an accelerometer, a motion capture system and a force sensor. We compose a matrix from the the time-series data and use the left singular vectors of the matrix as the patterns of the motion and the singular values as a scalar, by which each corresponding left singular vector affects the matrix. Two experiments were conducted to testify the method. One is a gesture recognition experiment in which we categorize gesture motions by two kinds of models with the indexes of similarity and estimation using left singular vectors. The other is an ambulation evaluation experiment in which we distinguished the levels of walking disability using a 3D hyperplane constructed by the singular values. Finally we discuss the characteristic and significance of the embodied knowledge extraction using singular value decomposition proposed in this paper.

Walking is a fundamental human ability necessary for everyday life. We have developed an omnidirectional walker (ODW) for walking support to those who have walking disabilities. It is necessary for the ODW to know which direction the user is intending to go during walking support. A novel interface is proposed for the ODW to recognize directional intention according to the user's forearm pressures which are measured by force sensors embedded in the armrest. The relationship between forearm pressures and directional intention was extracted as fuzzy rules and an algorithm is proposed for directional intention identification based on distance type fuzzy reasoning method. In this paper, we conduct walking support experiments with the proposed method. The results show that the algorithm is applicable to directional control in walking support.

Walking is a vital exercise for health promotion and a fundamental ability necessary for everyday life. In the authors' previous studies, an omni-directional walker (ODW) has been developed for walking rehabilitation and walking support. In the case of walking support, it is necessary for the ODW to know which direction its user is intending to go according to the user's manipulation. However, since the directional intention and physical manipulation of a user are not always consistent with each other, it is an important subject to identify the real directional intention from physical manipulation. In this paper, a method is proposed to recognize a user's directional intention according to the pressures, which are measured by sensors embedded in the ODW's armrest, from the user's forearms. Firstly, the relationship between forearm pressure and directional intention was extracted as fuzzy rules. Then an algorithm is proposed for directional intention identification based on Distance-Type Fuzzy Reasoning Method (DTFRM). Finally, the effectiveness of the algorithm is verified by experiments, which show that reasoning results are consistent with the intended directions.