Search

Purpose This paper aims to investigate whether electroencephalography (EEG) technology is effective in qualifying the tactile sensation evoked by non-steady cutaneous electrical stimulation. EEG is a novel method for electrotactile analysis and has demonstrated the discrimination ability for electrotactile sensation under steady contact conditions in recent years. However, in non-steady contact conditions, it is necessary to test its effectiveness. This study aims to explore an objective analysis method in comparison to psychophysical approach and to provide a methodology for non-steady electrotactile research. Design/methodology/approach With EEG experimentation on 13 volunteers, the authors collected evoked potentials by the predesigned “1” and “0” stimulation events. In addition, with a series of data preprocessing including artifact elimination, band-pass filtering, baseline normalization, data superposition and fast Fourier transform transformation, the authors got the power spectrum of alpha, beta and gamma rhythms. Furthermore, statistics analysis and ANOVA test were adopted for exploring the discrepancy of the spectrum characterizations for different non-steady electrostimulation events. Findings The EEG power spectrum of the central cortical brain is valuable in discriminating the two types of stimulation events. The power of alpha rhythm especially in the central cortical brain evoked by event “1,” whose current level is equal to the threshold, was significantly lower than that evoked by event “0,” whose level is less than the threshold (p < 0.05). Then, the power of the beta rhythm presented counter-change (p < 0.05). This study suggests that EEG may have the potential to qualify non-steady electrotactile sensation for engineering applications. Research limitations/implications Limiting factors of non-steady electrotactile stimulation were considered in this study. Different tapping frequency and contact time should be investigated in future studies. Originality/value This paper fulfills a challenge in qualifying the tactile sensations evoked by non-steady electrical stimulation with EEG characteristics.

The wires with chemical composition Ni20Mn20Ga20Gd20Co20 were prepared by hot-magnetic drawing and the microstructure evolution characteristics, martensitic transformation and MFIS process were investigated in detail, respectively. The results showed that a multiphase structure with γ phase and martensite was observed in samples when the magnetic field was 0 T to 0.2 T during the hot-magnetic drawing process. With the magnetic field increased to 0.5 T, due to the atomic diffusion by severe thermoplastic deformation and high external magnetic field, a single-phase structure with L10 type twin martensite was found in the sample. Moreover, an obvious increasing trend in martensitic transformation temperature in the sample was found by the enhancement of the magnetic field during the hot-magnetic drawing process. The highest phase transition temperature rose to about 600 °C when the magnetic field reached 0.5 T. Finally, the property of SME and MFIS in the sample can be enhanced by the magnetic field increasing during the hot-magnetic drawing process, excellent performance of SME was obtained at low total strain, and MFIS was achieved at 4.47% at a magnetic field of 8007 Oe in the sample in the 0.5 T magnetic field during the hot-magnetic drawing process.

In this paper, we propose a real-time gesture recognition model based on a feedforward artificial neural network for surface electromyography (sEMG) signals, which contains four processes: pre-processing, feature extraction, classification, and post-processing. In the feature extraction and selection stage, sliding windows are used to segment the data, and time-domain features and convolution results are used as input features. To avoid over-fitting of training, a dropout layer is added to the neural network model. In the post-processing part, two statistical methods are used: majority voting and continuous voting result verification. Finally, a test trial was performed using sEMG signals collected from the human body. The results show that the algorithm can achieve 92.7% recognition accuracy and has good real-time performance.

The effect of a high-entropy design on martensitic transformation and magnetic field-induced strain has been investigated in the present study for Ni-Mn-Ga-Co-Gd ferromagnetic shape-memory alloys. The purpose was to increase the martensitic transition temperature, as well as the magnetic field-induced strain, of these materials. The results show that there is a co-existence of β, γ, and martensite phases in the microstructure of the alloy samples. Additionally, the martensitic transformation temperature shows a markedly increasing trend for these high-entropy samples, with the largest value being approximately 500 °C. The morphology of the martensite exhibits typical twin characteristics of type L10. Moreover, the magnetic field-induced strain shows an increasing trend, which is caused by the driving force of the twin martensite re-arrangement strengthening.

As explosive growth of mobile data traffic brings great challenges to the mobile networks, how to reasonably deploy the latest generation mobile networks to meet the needs of users for communication becomes an urgent technical problem to be solved. Since machine learning technology has shown its superiority in processing big data in recent years, researches on the application of machine learning technology in wireless communication are expanded and explored gradually. In this paper, a wireless signal propagation model based on deep learning neural network is proposed to evaluate the wireless signal propagation characteristics and signal coverage status in different scenarios in the complex and changeable residential outdoor activity areas. The massive data generated in practical applications are reasonably utilized to predict the signal propagation characteristics of each cell with high accuracy and stability. The model has is of good adaptability to different complex scenarios and can providing provide a certain reference value and engineering guidance for further wireless network deployment and optimization development.

To precisely formulate the color during the production of gelatin hollow capsules, a gelatin hollow capsule color matching model based on particle swarm optimization back propagation (PSO-BP) neural network is proposed. The experiment of gelatin color matching was carried out and images of the gelatin solution were collected. MATLAB image processing toolbox was utilized to conduct image filtering and image segmentation, improving the accuracy of color characteristics extracting. The BP neural network’s initial parameters were optimized by the particle swarm optimization algorithm, building the relation model between the RGB values of two sets of gelatin solution and the difference between the added pigments. The experimental data was used to train and test the model. It can be concluded that the prediction effect of the model is satisfying, and the training speed and prediction accuracy have been improved to a certain extent compared with the model of traditional BP neural network.

In recent years, the application of automated guided vehicles (AGVs) in industrial manufacture has attracted more and more attention. Based on reinforcement learning, an automated guided logistics handling vehicles routing algorithm was proposed under a multi-task scenario. First, the algorithm calculated routing paths for each task node using Q-learning to solve the “curse of dimensionality” problem. Then, with the objective of comprehensive cost considering the time window, the optimal routing sequence was obtained using the depth first search algorithm (DFS). Finally, an example is given to illustrate the algorithm.

In the field of fast and large-scale image retrieval, the binary hash code based on convolutional features enjoys powerful advantages. CNN model's leading recognition ability shows that its high-level features distribute intensively and have stable between-class distances. On the basis of the distribution characteristic of convolutional features, this paper presents a new method of improving retrieval performance by means of using the category centers of high-level features to enhance the common semantic content of high-level features of the query image and making the consequent binary hash code gain better retrieval results. Experimental results indicate that (i) the category centers of CNN feature offer comparable discriminability with the initial model when identifying the semantic information of images; (ii) before calculating binary hash code, the integration of the CNN feature of a query image with category centers weighted by predicted probabilities can boost performance in similar image retrieval. State-of-the-art retrieval performance is obtained on the Cifar-10 dataset in this paper among all hash methods.

In recent years, deep convolutional networks have made a milestone progress in the field of image recognition. However, the recognition ability of the deep convolution network declines in the case of unbalanced training data, in terms of categories with fewer training images, the recognition ability of the deep convolution network declines more seriously. Aiming at this kind of problem, this paper presents a new classification method which recognizes a query image by comparing distances between category centers of CNN features of the whole training dataset and the corresponding CNN feature of this query image. The experimental results on Cifar-10 and Cifar-100 show that the claimed method can more accurately identify images whose categories have only a few training samples and that the mean precision of the recognition can be improved effectively.

The Co35Ni32Al32Dy bulk alloys were prepared via suction casting technique and the microstructures obtained after furnace cooling and rapid solidification (RS) were investigated, respectively. The results showed that the microstructure of the alloy that contained the β and γ phases and the intermetallic compound Co5Dy transformed into martensite and γ dual-phase, when the cooling rate (RC) under furnace cooling was increased to 4000 K s−1. Moreover, the dendritic type γ phase fragmented into smaller pieces and distributed with in the grain, and the solid solubility of the alloying element Dy in the matrix was significantly improved by RS. The mechanical properties were substantially enhanced by the RS technique and the maximum tensile strength of 3395 MPa, with the elongation of 35.8% was obtained for the alloy with the cooling rate of 4000 K s−1. Finally, the fracture surface of the sample after furnace cooling and RS demonstrated typical characteristic of ductile fracture. With the increase in RC value, the dimples on fracture surface became more uniform in size with homogeneous distribution.

Conventional edge-based object pose estimation in robot vision shows low accuracy and slow convergence performance. In this paper, a corner-based object pose estimation method is proposed and studied. The classical pinhole camera model and model-based object pose estimation or tracking are employ, as well as the iterative optimization for accurate estimation. In our method, a new corner matching strategy is proposed since good corresponding is easy to be met for image corners. Then, good fitting results between the CAD model and object in the images can be obtained. Experimental results show that the new method can achieve more accurate and faster pose estimation than the conventional methods.

Although various stair climbing or cleaning robots have been proposed and developed, few of them can move on the tread to do cleaning. Hence, the avoidance of falling and collision for a stair-cleaning robot has not been thoroughly solved. In this paper, based on our newly developed stair-cleaning robot, several methods are proposed or introduced to solve three kinds of problems concerning the avoidance of falling and collision. The proposed methods solve the problem of no falling before and when the robot is climbing. The proposed methods can assure no falling and few slight collisions when the robot is moving on the tread. A new turning approach is proposed and solves the problem of successful turning on the tread of the stairs for a stair-cleaning robot. Experimental results show that our methods are effective and stable.

Various stair-climbing or cleaning mechanisms have been proposed or developed in the past decades. Most of them are big-sized, complex, and/or expensive, which hinder their practical application. In this study, a new stair-cleaning robot is proposed and developed. In the new robot, a pair of legs is used for stair climbing. Rotating around the main body, the legs of the robot drive the feet to execute revolving but translational motions, thereby allowing the robot to climb. The new robot is highly compact compared with most traditional stair-cleaning or climbing robots, because only one motor is utilized to drive the climbing mechanism and the legs can be retracted at both sides of the robot when it is not climbing. Its compact structure enables the new robot to move along the riser to do cleaning on the stairs, which is a prominent advantage over most similar robots. The cleaning device is designed similar to that of current ground cleaning robots. The sensing and control system is designed for successful stair climbing and avoidance of falls and collisions. Experiments show that the new robot has high success rates in climbing up or down common types of stairs, and thus can do cleaning on the related stairs.

ELM algorithm has been widely applied in gesture recognition. When the dataset is multi-objective, however, using classical ELM algorithm directly may produce a big recognition error. To address this problem, an improved ELM recognition algorithm is proposed. The presented ELM algorithm is characterized as building separated ELM network for each gesture instead of constructing a unified ELM network for all gestures. A simplified and optimized feature is proposed. Comparison experiment between the classical ELM algorithm and the optimized ELM algorithm aiming to four classical gestures are conducted. The result shows that the training accuracy of the improved algorithm is about 5.25 times of the classical algorithm, and the right recognition ability of the improved algorithm is more than 1.8 times than the classical algorithm. The training time of the optimized algorithm is less than that of the classical algorithm.

In automatic 3D reconstruction, determining next best view (NBV) is the key problem for continuous automatic view planning. The mass vector chain (MVC) is one of the classical methods, but it can only determine the NBV direction. A visual servo-based approach is proposed in this paper for determining the orientation and direction of NBV. In the proposed approach, the 2-1/2-D visual servoing is used and modified in which MVC method is employed for orientation control and the image-based visual servoing is employed for position control. The MVC method is used in an open-loop, which decreases the dimensionality of the image Jacobian. The experimental results show that the proposed method can guide the robot to the desire views automatically. The final 3D reconstruction surface of an object is basically complete, only with a few tiny blind areas due to too many concave small surface patches of the object.

Conventional D-H model based inverse kinematic solution algorithms have to hypothesize that d 6 =0 at the end of the robot hand, or the joint variables of D-H based kinematic equations can hardly be isolated to be solved. In practical application, d 6 ≠ 0 usually occurs, hence it is required for supplemental transformation between the last frame of the D-H models and real operation frame, which decreases the efficiency. In this paper, a new inverse kinematic solution without the hypothesis of d 6 =0 is proposed. We found that the joint variables of the partially decoupled robot can be separated without great difficulty even though d 6 ≠ 0. The new approach uses a new multiplication sequence of the inverse matrix for solving the variables of the kinematic equation of the robot. The inverse solution process is straightforward and computationally cheap. Last, some exemplifications are given to valid the accuracy of the proposed approach.

Aiming at the showing art of cartoon animation, a new art form based on tactile technology is suggested to render the cartoon animation action in this paper. The principle of tactile cartoon animation is introduced, and some tactile illusion phenomena which supporting the tactile cartoon animation are presented too. On this basis, the rendering techniques of tactile cartoon animation are discussed, and finally, three cartoon animation rendering modes based on the movement and synthesis of basic elements are suggested, which are basic movement, static built-up and dynamic built-up.

Aiming at the cultural relic exhibition problem in virtual museum, the tactile rendering technology for epigraph and inscription exhibition is suggested. With the tactile display technology, one can not only see the real relic, hear explication, but also touch the cultural relic with his finger when he explores the virtual museum. The principle of the virtual tactile rendering system is illustrated, and both the tactile glove and the bar frame structure are designed too, and then the vision-tactile image process algorithms are introduced, finally, two experiments on epitaph and inscription are demonstrated. Experiments show that operators can perceive and recognize the character information on the epitaph and inscription surface with tactile technology successfully.

Pressure stimulus causes skin deformation and tactile sensation on the fingertip. Both theoretical approach and experimental technique may be used to investigate the relationship between the deformation and the sensation. Building an appropriate skin model is the most important step for further theoretical and experimental analysis. In this paper, a two dimensional (2D) fingertip biomechanical model employing finite element (FE) method is proposed based on the physiological structure of skin. With biomechanical and electrophysiological simulations, the predicted distributions of the strain energy density (SED) and the stress/strain are obtained. The relation between the predicted biomechanical responses of the subcutaneous tissue and the discharged rate reported in the literatures are investigated. The results show that the soft tissues of fingertips are very sensitive to the external stimulus, and the spatial distribution characteristics of SED within soft tissues can explain the evoked charging rate of mechanoreceptors effectively. The simulation data of the proposed FE model is highly consistent with the verified data.

For investigating the effects of stimuli on the deformations within the soft tissues of fingertips and the dependence of the tactile sensation on the deformations under pressure, in this paper, a finite element model is developed, to simulate the procedure of a fingertip touching a sharp wedge. Characteristics of the strain energy density (SED) distribution within the soft issues are analyzed. Simulation results show that the soft tissues of fingertips are very sensitive to stimuli, and the spatial distribution characteristics of strain energy density within soft tissues can best explain the evoked charging rate of mechanoreceptors.