순환 신경망을 이용한 보행단계 분류기

Soil types mapping and the spatial variation of soil classes are essential concerns in both geotechnical and geoenvironmental engineering. Because conventional soil mapping systems are time-consuming and costly, alternative quick and cheap but accurate methods need to be developed. In this paper, a new optimized multi-output generalized feed forward neural network (GFNN) structure using 58 piezocone penetration test points (CPTu) for producing a digital soil types map in the southwest of Sweden is developed. The introduced GFNN architecture is supported by a generalized shunting neuron (GSN) model computing unit to increase the capability of nonlinear boundaries of classified patterns. The comparison conducted between known soil type classification charts, CPTu interpreting procedures, and the outcomes of the GFNN model indicates acceptable accuracy in estimating complex soil types. The results show that the predictability of the GFNN system offers a valuable tool for the purpose of soil type pattern classifications and providing soil profiles.

The present work was aimed to assess the variation of grip strength with the changes of posture and body joint angles and to establish an optimal body posture and joint angle for the maximum grip strength for adult Bengalee population. Methods: Adult male (n=156) and female (n=224) were involved in this study. The hand grip strength was measured by using Lafayette hand grip dynamometer. Results: The results revealed that male subjects had a significantly greater (p<0.001) grip strength compared to that of female subjects and the highest grip strength was found in standing posture with elbow angle of 900 in both adult male and female subjects. Conclusions: From the results it may be recommended that the grip strength of Bengalee population may be measured in standing posture with elbow angle of 90degrees, shoulder angle of 1800, and the trunk and wrist in neutral positions to provide maximum force.

A neural network approach to modeling fuzzy preference relations for multiple criteria decision making

Following the idea of using first-order time derivatives, this paper presents a general recurrent neural network (RNN) model for online inversion of time-varying matrices. Different kinds of activation functions are investigated to guarantee the global exponential convergence of the neural model to the exact inverse of a given time-varying matrix. The robustness of the proposed neural model is also studied with respect to different activation functions and various implementation errors. Simulation results, including the application to kinematic control of redundant manipulators, substantiate the theoretical analysis and demonstrate the efficacy of the neural model on time-varying matrix inversion, especially when using a power-sigmoid activation function.

Wheelchair propulsion: Force orientation and amplitude prediction with Recurrent Neural Network

A new hybrid type of recurrent neural network is reported in this paper. Unlike those of the general recurrent neural networks, the activation functions of its hidden neurons are chosen to be Gaussian functions. The performance of this hybrid network for nonlinear nonstationary time series prediction is tested. Two popular time series produced by the Mackey-Glass equation and the Lorenz equation are used in the simulation.

Exponential stabilization and synchronization for fuzzy model of memristive neural networks by periodically intermittent control

With the improvement of computer performance, virtual reality (VR) as a new way of visual operation and interaction method gives the automatic lip-reading technology based on visual features broad development prospects. In an immersive VR environment, the user’s state can be successfully captured through lip movements, thereby analyzing the user’s real-time thinking. Due to complex image processing, hard-to-train classifiers and long-term recognition processes, the traditional lip-reading recognition system is difficult to meet the requirements of practical applications. In this paper, the convolutional neural network (CNN) used to image feature extraction is combined with a recurrent neural network (RNN) based on attention mechanism for automatic lip-reading recognition. Our proposed method for automatic lip-reading recognition can be divided into three steps. Firstly, we extract keyframes from our own established independent database (English pronunciation of numbers from zero to nine by three males and three females). Then, we use the Visual Geometry Group (VGG) network to extract the lip image features. It is found that the image feature extraction results are fault-tolerant and effective. Finally, we compare two lip-reading models: (1) a fusion model with an attention mechanism and (2) a fusion model of two networks. The results show that the accuracy of the proposed model is 88.2% in the test dataset and 84.9% for the contrastive model. Therefore, our proposed method is superior to the traditional lip-reading recognition methods and the general neural networks.

The purpose of the present study was to compare the body configurations and the joint function during standing long jump in 6-yr-old children and adult males. Twelve healthy adult males and eight (one male and seven females) 6-yr-old kindergartners participated in this study. Subjects performed standing long jump on a force platform with full effort. Body segment and joint angles were analyzed by high speed videography (100 frames.s-1). Using kinetic and kinematic data, joint moments, power, and work done were calculated through a free body diagram. Average standing long jump performances were +1.5 SD above Japanese norm in both adults and children. A wide range of motion of the lower limb segments during flight phase was found in the adults. In the crouch prior to take off, joint muscle power peaks appeared in the same order, and joint contribution to the total work done showed almost the same values in both adults and children. It is suggested that gross motor pattern before take off is almost accomplished by the 6-yr-old. Judging by hip joint work, adult performance was much better than that of children, since the adults exerted relatively higher negative work from the onset of preparatory movement to the point of lowest center of gravity during crouch. It was concluded that skilled 6-yr-old children have not fully developed either hip negative work during preparatory movement or body configuration in the flight phase, although gross motor pattern before take off phase was accomplished to skilled adult level.

Train running safety is considered one of the key criteria for advanced highway trains and bogies. While a number of existing research studies have focused on its measurement and monitoring, this study proposes a new and effective train running a safety prediction framework. The wheel derail coefficient, wheel rate of load reduction, and wheel lateral pressure are considered the decision variables for the safety framework. Data for actual measured rail conditions and vibration-based signals are used as the input data. However, advanced trains and bogies are influenced more by their inertial structures and mechanisms than by railway conditions and external environments. In order to reflect their inertial influences, past data of output variables are used as recurrent data. The proposed framework shares advantages of a general deep neural network and a recurrent neural network. To prove the effectiveness of the proposed hybrid deep-learning framework, numerical analyses using an actual measured train-railway model and transit simulation are conducted and compared with the existing deep learning architectures.

Recurrent neural networks: A constructive algorithm, and its properties

Robots are widely used in various engineering fields, and the solution to their trajectory tracking problem has attracted increasing attention. Such a problem can be typically transformed into a time-varying nonlinear equation (TVNE). For complex and high-precision robot trajectory tracking problems, a fast and low-error tracking solution is necessary. Therefore, a varying-parameter recurrent neural network (VPRNN) model with a modified power-type time-varying parameter is proposed for solving TVNE. An improved sign-bi-power function is selected for the activation function, then the VPRNN model achieves fixed-time convergence. Numerical comparisons with the general fixed-parameter recurrent neural network model are performed, which demonstrates the superiority of our VPRNN model. Besides, the proposed VPRNN model is successfully used to solve the trajectory tracking problem of a three-link robot, which shows its feasibility in practical applications.KeywordsVarying-parameter recurrent neural networkTime-varying nonlinear equationFixed-time convergenceRobot manipulator control

The redundant manipulator as a ubiquitous and essential component of robots, acts as a momentous role in mechanized production. The motion control of a redundant manipulator is an essential problem that must be handled. Consequently, for the repetitive motion control of redundant manipulators, a recurrent neural network (RNN) model for the double-index (DI) scheme is researched in this paper. The DI scheme sets the minimum kinetic energy (MKE) and minimum joint-angle offset (MJAO) as optimization indexes, for which the time-dependent weights are designed, and takes the dynamic equations of redundant manipulators and joint limit constraints into consideration as well. Then, the DI scheme is reformulated as a quadratic programming (QP) problem. Besides, the RNN model based on the iteration of related parameters is derived to solve the QP problem, through which joint data (i.e., joint angles, joint velocities, and joint accelerations) are obtained to drive the motion of the redundant manipulator. To verify the effectiveness of the DI scheme for motion control of redundant manipulator, simulation results of the DI scheme solved by the RNN model, the minimum acceleration norm (MAN) scheme and MKE scheme are compared.

In this paper, we propose a general projection neural network for solving a wider class of optimization and related problems. In addition to its simple structure and low complexity, the proposed neural network include existing neural networks for optimization, such as the projection neural network, the primal-dual neural network, and the dual neural network, as special cases. Under various mild conditions, the proposed general projection neural network is shown to be globally convergent, globally asymptotically stable, and globally exponentially stable. Furthermore, several improved stability criteria on two special cases of the general projection neural network are obtained under weaker conditions. Simulation results demonstrate the effectiveness and characteristics of the proposed neural network.

&lt;abstract&gt;&lt;p&gt;This paper is concerned with piecewise pseudo almost periodic solutions of a class of interval general BAM neural networks with mixed time-varying delays and impulsive perturbations. By adopting the exponential dichotomy of linear differential equations and the fixed point theory of contraction mapping. The sufficient conditions for the existence of piecewise pseudo almost periodic solutions of the interval general BAM neural networks with mixed time-varying delays and impulsive perturbations are obtained. By adopting differential inequality techniques and mathematical methods of induction, the global exponential stability for the piecewise pseudo almost periodic solutions of the interval general BAM neural networks with mixed time-varying delays and impulsive perturbations is discussed. An example is given to illustrate the effectiveness of the results obtained in the paper.&lt;/p&gt;&lt;/abstract&gt;