Dexterous control of robotic hand driven by human motor intent has drawn a lot of attention in both industrial and rehabilitation scenarios. Providing simultaneous and proportional control has become a prevailing solution recently. Towards improving the finger kinematics estimation precision and reducing its computational cost, a convolution model with attention mechanism (CNN-Attention) was proposed in this study. For comparison purpose, two previously used deep learning models, the long short-term memory (LSTM) and the Sparse Pseudo-input Gaussian processes (SPGP) were also included. By using surface electromyography (sEMG) and kinematic signals corresponding to six hand grasp movements, the estimation performance of each of the three models was evaluated with three measures, Pearson Correlation Coefficient (CC), Root Mean Square Error (RMSE), and coefficient of determination (R2) between real and estimated joint angles. The results demonstrated that the proposed CNN-Attention model outperformed LSTM and SPGP significantly ( p -value<0.05), with an average value of CC, RMSE, and R2, 0.87, 9.65 degrees, and 0.73, respectively. Also, the CNN-Attention model is more stable and versatile over various subjects and joint angles in comparison to other models. Additionally, the computational time to build a CNN-Attention was obviously shorter than that to train a LSTM model (43.00 ± 4.25 min vs. 73.40 ± 5.81 min). These findings suggest that the CNN-Attention would be a promising model for continuous estimation of hand movements in the human-machine interaction and cooperation.

Hyperactive reflexes are commonly observed in patients with spinal cord injury (SCI) but there is a lack of convenient and quantitative characterizations. Patellar tendon reflexes were examined in nine SCI patients and ten healthy control subjects by tapping the tendon using a hand-held instrumented hammer at various knee flexion angles, and the tapping force, quadriceps EMG, and knee extension torque were measured to characterize patellar tendon reflexes quantitatively in terms of the tendon reflex gain (Gtr), contraction rate (Rc), and reflex loop time delay (td). It was found that there are significant increases inGtrandRcand decrease intdin patients with spinal cord injury as compared to the controls (P<0.05). This study presented a convenient and quantitative method to evaluate reflex excitability and muscle contraction dynamics. With proper simplifications, it can potentially be used for quantitative diagnosis and outcome evaluations of hyperreflexia in clinical settings.

Under a radar-based mini-size unmanned aerial vehicle (mini-UAV) classification scenario, the mini-UAV physical parameters play an important role. The parameters can be retrieved from radar blade flashes induced by rotor blades. In this work, we propose a novel method for enhancing the distinctive-ness between blade flashes so that the physical parameters can be better estimated. Essentially, the radar micro-Doppler signatures (m-DS) is decomposed using a time-frequency analysis method, empirical mode decomposition (EMD). We then reconstruct the blade flashes using the selected blade-flash-characterized mode functions. From the spectrogram, the reconstructed blade flashes can be better seen than without using EMD.

Background: In the process of fracture reduction, there are some errors between the actual trajectory and the ideal trajectory due to mechanism errors, which would affect the smooth operation of fracture reduction. To this end, based on self-developed parallel mechanism fracture reduction robot (FRR), a novel method to reduce the pose errors of FRR is proposed. Methods: Firstly, this paper analyzed the pose errors, and built the model of the robot pose errors. Secondly, mechanism errors of FRR were converted into drive bar parameter’s errors, and the influence of each drive bar parameter on the robot pose error were analyzed. Thirdly, combining with Cauchy opposition-based learning and differential evolution algorithm (DE), an improved whale optimization algorithm (CRLWOA-DE) is proposed to compensate the end-effector’s pose errors, which could improve the speed and accuracy of fracture reduction, respectively. Results: The iterative accuracy of CRLWOA-DE is improved by 50.74%, and the optimization speed is improved by 22.62% compared with the whale optimization algorithm (WOA). Meanwhile, compared with particle swarm optimization (PSO) and ant colony optimization (ACO), CRLWOA-DE is proved to be more accurate. Furthermore, SimMechanics in the software of MATLAB was used to reconstruct the fracture reduction robot, and it was verified that the actual motion trajectory of the CRLWOA-DE optimized kinematic stage showed a significant reduction in error in both the x-axis and z-axis directions compared to the desired motion trajectory. Conclusions: This study revealed that the error compensation in FRR reset process had been realized, and the CRLWOA-DE method could be used for reducing the pose error of the fracture reduction robot, which has some significance for the bone fracture and deformity correction.