Papers by Keyword: Radial Basis Function (RBF)

Abstract: A new method for premature ventricular contraction (PVC) detection and classification is presented. The proposed algorithm is constituted of two principal phases: the features extraction and reduction phase and the optimized classification phase. In the first phase, the discrete cosine transform (DCT) and the continuous wavelet transform (CWT) are applied on each ECG beat to generate an augmented features vector. For the optimized classification phase, the radial basis function (RBF) neural network classifier is trained and optimized by the bat algorithm. For the aim of performances evaluation of the proposed method, the MIT-BIH arrhythmia database has been used. Consequently, the BAT-RBF classifier yielded an overall sensitivity of 95,2% and an accuracy of 98,2%, confirming clearly the competitiveness of the proposed method compared to some recent and powerful algorithms.

Abstract: In present paper, deflection and stress of laminated composite plates are analyzed by a meshless local collocation method based on inverse multiquadrics radial basis function. This method approximates the governing equations based on first-order shear deformation theory using the nodes in the support domain of any data center. Transverse displacement, normal stresses, and shear stresses of the simply supported laminated composite plates under sinusoidal load are computed by the present method. The convergence characteristics are studied by several numerical examples. The present results are compared with available published results which demonstrate the accuracy and efficiency of present method.

Abstract: RBF neural network have advantages of training simple, fast efficiency of learning, easy to fall into local minima, etc..It is widely used to solve the problem in signal processing and pattern recognition. Although the common RBF network is relatively easy to build, but because of the structure is usually fixed or high complexity, resulting in learning time is too long or network resource waste. For these reasons, proposed using extended Kalman filter as the RBF learning algorithm, and using double radial function in the hidden layer. By approaching the basis of the results of the analysis clearly shows that the network model than the other categories have a stronger generalization.

Abstract: Prediction of suspended sediment concentration in hyper-concentrated rivers is a crucial task in modeling and designing of hydraulic structures such as dams, reservoirs, barrages and water intake inlets. In this study, suspended sediment concentration in Kinta River has been predicted using radial basis function (RBF) neural network modeling technique. Time series of suspended sediments and stream discharge data from 1992 to 1995 are used in the training and testing stages of the model. The data were divided into two sections based on the model stages as training and testing. The Thin Plate Spline (TPS) basis function was used to establish TPS - RBF prediction model. The input neurons were selected based on previous studies about the suspended sediment prediction models. The number of hidden neurons was determined by trial and error method. The spread of the basis function was determined by normalization method. The performance of the prediction model was evaluated using three statistical performance measures namely root mean square error (RMSE), coefficient of efficiency (CE) and coefficient of determination (R²). The results showed that the TPS – RBF model predicted the suspended sediment values close to the observed data. The statistics of the model showed that the prediction model performed very well and produced R² values close to one in both training and testing stages.

Abstract: The recent growing interest for indoor localization-based services has created a need for more accurate and real-time indoor localization solutions. Indoor localization based on existing WiFi signal strength is becoming increasingly prevalent and ubiquity. In this paper, we utilize the information of the signal strength received from the surrounding access points (APs) to determine the user localization. The propose algorithm based on support vector machines (SVM) algorithm, and comparing with three kernel functions, radial basis function (RBF) performs best of all. Experimental results indicate that the proposed algorithm leads to improvement on localization accuracy.

Abstract: This paper adopts the Radial Basis Function (RBF) Neural Networks to conduct a spatial prediction on the mercury pollution situation of the Jiapigou gold mine area, locate the primary pollution sources, delineate the pollution area according to the mercury concentration data of 27 soil samples from this area, and draws the mercury concentration isoline with the gridded data. Compared with the methods in the past such as classical statistics and BP Neural Networks to analyse the soil pollution, this method presents advantages such as the quantification of the result, the explicitness of the pollution area, and the ability to explain the blind area of the samples.

Abstract: Simulation of transient bioheat transfer in a two dimensional (2D) human eye model is conducted using a newly developed hybrid fundamental solution-finite element method (HFS-FEM) coupling with the radial basis function (RBF) approximation. Firstly, a time stepping scheme based on the finite difference method (FDM) is used to handle time variable in the transient Pennes bioheat equation. Secondly, the particular solution of the governing equation is approximated by a RBF approach. Then, the homogeneous solution is calculated by means of HFS-FEM. The obtained results are compared with those from ABAQUS and a good agreement between them is observed.

Abstract: Artificial intelligent techniques are being actively applied in many applications. With their powerful learning capability of neural networks and reducing the optimizing search space by prior knowledge rules of Fuzzy systems have been proven to be rather efficiency. In this research, the hybrid Neuro-Fuzzy system (NF) is proposed to be utilized as a predictor of the Grade Point Average (GPA) of students for future planning where the Radial Basis Function (RBF) is implemented as a neuro-fuzzy system. The NFs parameters consisted of centre and width of the Gaussian membership function and weight between input layer and output layer are automatically tuned by using Genetic Algorithms (GA) referred as NF-GA. The collected data is then tested and trained through NF-GA system with Minimum Mean Square Error (MMSE) technique. It has been shown that our proposed model is capable of prediction GPA by accurately 93%.The performance comparison between the proposed NF-GA and Multiple Regression Analysis (MRA) gives performance significantly by reducing the average error of the prediction down to 10%. Keywords: Neuro-fuzzy system, Genetic Algorithms, Multiple Regression Analysis, Radial Basis Function.

Abstract: Maximum security is essential for a robotic device to achieve its optimum control. In this research, we present robotic motions, controlled by technological speech recognition techniques, using commands of Thai Speech Recognition (TSR). Examples of such speech words used are; Sai, Khwa, Na, Lang Khun, Long and Yood, which are the equivalent English language representations for; turn left, turn right, forwards, backwards, upwards, downwards, and stop, respectively. The speech commands are independent for any particular user and so, as a result, they are highly beneficial for general and practical use. The three main important parts of this paper comprise; Pre-processing, Discrete Fourier Transform (DFT), and Back Propagation Neural Network (BPN). The experimental results, when reviewed, exhibit results showing that the average accuracy percentages are equal to 71.00% for female commands, and 70.00% for male commands, respectively.

Abstract: To obtain the improvement of analog circuit fault diagnosis, a RBF diagnosis model based on an Adaptive Genetic Algorithm (AGA) is proposed. First an adaptive mechanism about crossover and mutation probability is introduced into the traditional genetic algorithm, and then AGA algorithm is used to optimize the network parameters such as center, width and connection weight. The experiment simulation indicates that the proposed model has exact diagnosis characteristic.