An ideal hand prosthesis should provide satisfying functionality based on reliable decoding of the user's intentions and deliver tactile feedback in a natural manner. The absence of tactile feedback impedes the functionality and efficiency of dexterous hand prostheses, which leads to a high rejection rate from prostheses users. Thus, it is expected that integration of tactile feedback with hand prostheses will improve the manipulation performance and enhance perceptual embodiment for users. This paper reviews the state-of-the-art of non-invasive stimulationbased tactile sensation for upper-extremity prostheses, from the physiology of the human skin, to tactile sensing techniques, noninvasive tactile stimulation, and an emphasis on electrotactile feedback. The paper concludes with a detailed discussion of recent applications, challenging issues, and future developments.

It is evident that the dominant therapy of functional electrical stimulation (FES) for stroke rehabilitation suffers from heavy dependency on therapists experience and lack of feedback from patients status, which decrease the patients' voluntary participation, reducing the rehabilitation efficacy. This paper proposes a closed loop FES system using surface electromyography (sEMG) bias feedback from bilateral arms for enhancing upper-limb stroke rehabilitation. This wireless portable system consists of sEMG data acquisition and FES modules, the former is used to measure and analyze the subject's bilateral arm motion intention and neuromuscular states in terms of their sEMG, the latter of multi-channel FES output is controlled via the sEMG bias of the bilateral arms. The system has been evaluated with experiments proving that the system can achieve 39.9 dB signal-to-noise ratio (SNR) in the lab environment, outperforming existing similar systems. The results also show that voluntary and active participation can be effectively employed to achieve different FES intensity for FES-assisted hand motions, demonstrating the potential for active stroke rehabilitation.

Fine multi-functional prosthetic hand manipulation requires precise control on the pinch-type and the corresponding force, and it is a challenge to decode both aspects from myoelectric signals. This study proposed an attribute-driven granular model (AGrM) under a machine learning scheme to solve this problem. The model utilises the additionally captured attribute as the latent variable for a supervised granulation procedure. It was fulfilled for EMG-based pinch-type classification and the fingertip force grand prediction. In the experiments, sixteen channels of surface electromyographic signals (i.e. main-attribute) and continuous fingertip force (i.e. sub-attribute) were simultaneously collected while subjects performing eight types of hand pinches. The use of AGrM improved the pinch-type recognition accuracy to around 97.2% by 1.8% when constructing eight granules for each grasping type, and received more than 90% force grand prediction accuracy at any granular level greater than six. Further sensitivity analysis verified its robustness with respect to different channel combination and interferences. In the comparison with other clustering-based granulation methods, AGrM achieved comparable pinch recognition accuracy but was of lowest computational cost and highest force grand prediction accuracy.

encoding of electrical stimulation. In future research, the design of a general electrotactile feedback enhanced hand rehabilitation platform with a standardised stimulation parameter optimisation will be addressed and further validated on the subjects with limbimpairments and amputation.

It is of great importance to decode motion dynamics of the human limbs such as the joint angle and torque in order to improve the functionality and provide more intuitive control in human-machine collaborative systems. In order to achieve feasible prediction, both the surface electromyography (sEMG) and A-mode ultrasound were applied to detect muscle deformation and motor intent. Six abled subjects were recruited to perform five trails elbow isokinetic flexion and extension, and each trail contained five repetitions, with muscle deformation and sEMG signals recorded simultaneously. The experimental datasets were categorized as: the ultrasound-EMG combined datasets, ultrasound-only datasets and EMG-only datasets. The support vector machine (SVM) regression model was developed for both elbow joint angle and torque prediction, based on the above three kinds of datasets. The root-mean-square error (RMSE) and the correlation coefficients (R) were applied to evaluate the prediction accuracy. The results across all the subjects for different datasets indicated that the combined datasets and the ultrasound datasets were superior to the sEMG datasets both on elbow joint angle and torque prediction, and there were no significant differences between the combined datasets and the ultrasound datasets. It turns out that elbow angle and torque can be reconstructed by Amode ultrasound, and the significant findings pave the way towards the application of musculature-driven human-machine collaborative systems.Angle, torque, surface electromyography (sEMG), ultrasound, support vector machine (SVM), regression, isokinetic contraction.