International Journal of Humanoid Robotics

Tactile feedback is beneficial to improve the hand prosthesis performance, alleviate phantom pain, reduce muscle fatigue, etc. During the manipulation process, muscle fatigue not only causes discomfort to prosthesis users but also disturbs the surface electromyographic (sEMG)-based motion recognition, which significantly deteriorates the prosthesis functional performance. Efforts have been made to explore appropriate signal processing algorithms which could be less influenced by muscle fatigue. However, few studies concern how to alleviate muscle fatigue directly. Thus, this study proposes a novel method to avoid excessive muscle fatigue based on electrotactile feedback. A potable electrotactile stimulator is developed with adjustable parameters, multiple channels and wireless communication. It is implemented in a virtual hand grasping platform driven by sEMG signals to investigate the impact of tactile feedback on muscle fatigue. Experimental results show a higher success rate of grasping with electrotactile feedback than that with no feedback. Moreover, compared with grasp in the no feedback condition, there is an observable decrease of sEMG intensity when grasping a heavy object with electrotactile feedback, despite a comparable performance on the light and medium objects in both feedback conditions. It indicates that tactile feedback helps to alleviate muscle fatigue caused by excessive muscle contraction, especially when large strength is needed.

The learning of inter-day representation of electromyographic (EMG) signals across multiple days remains a challenging topic and not fully accommodated yet. This study aims to improve the inter-day hand motion classification accuracy via convolutional neural network (CNN)-based data feature fusion. An EMG database (ISRMyo-I) was recorded from six subjects on 10 days via a low density electrode setting. This study investigated CNNs’ capability of feature learning, and found that the output of the first fully connected layer (CNNFeats) was a decent supplement feature set to the most prevalent Hudgins’ time domain features in combination with fourth-order autoregressive coefficients (TDAR). Through adding the automatically learned CNNFeats to the handcrafted TDAR feature set, both linear discriminant analysis (LDA) and support vector machine (SVM) classifiers received $>$3% accuracy improvement. Similarly, taking TDAR as additional input to the CNN improved the accuracy by $>$1% in the comparison with the basic CNN. Our results also demonstrated that the CNN approach outperformed conventional approaches when multiple subjects’ data were available for training, while traditional approaches were more adept at presenting motion patterns for single subject. A preliminary conclusion is drawn that substantial “common knowledge/features” can be learned by CNNs from the raw EMG signals across multiple days and multiple subjects, and thus it is believed that a pre-trained CNN model would contribute to higher accuracy as well as the reduction of learning burden.

Discrimination of surface textures using tactile sensors has attracted increasing attention. Intelligent robotics with the ability to recognize and discriminate the surface textures of grasped objects are crucial. In this paper, a novel method for surface texture classification based on tactile signals is proposed. For the proposed method, first, the tactile signals of each channel (X, Y, Z, and S) are decomposed based on empirical mode decomposition (EMD). Then, the intrinsic mode functions (IMFs) are obtained. Second, based on the multiple IMFs, the sample entropy is calculated for each IMF. Therefore, the multi-IMF sample entropy (MISE) features are obtained. Last but not least, based on the two public datasets, a variety of machine learning algorithms are used to recognize different textures. The results show that the SVM classification method, with the proposed MISE features, achieves the highest classification accuracy. Undeniably, the MISE features with the SVM method, proposed in this paper, provide a novel idea for the recognition of surface texture based on tactile perception.

In this paper, the sensorless force/position control problem is investigated for a general class of dynamic contact systems with both motion sensor noise and unknown kinetic friction by designing a force observer-based controller. Firstly, in order to suppress the effect of motion sensor noise, a dead-zone extended state observer (ESO) is introduced, and the contact force is estimated. Then, based on the force estimate, a controller is designed to realize force/position tracking control, where the parameters of the observer and controller are obtained by a linear matrix inequality (LMI) method. The sufficient conditions are provided to ensure the stability of the closed-loop system in terms of LMIs. Finally, a numerical simulation is carried out to illustrate the applicability and effectiveness of the proposed method.