Browsing by Author "Chen, Yon-Ping"

Now showing 1 - 20 of 20

An evolutionary constructive and pruning algorithm for artificial neural networks and its prediction applications
(2012-06-01)
Yang, Shih-Hung
;
Chen, Yon-Ping
;
電機工程學系
;
Department of Electrical and Computer Engineering
We propose a method for designing artificial neural networks (ANNs) for prediction problems based on an evolutionary constructive and pruning algorithm (ECPA). The proposed ECPA begins with a set of ANNs with the simplest possible structure, one hidden neuron connected to an input node, and employs crossover and mutation operators to increase the complexity of an ANN population. Additionally, cluster-based pruning (CBP) and age-based survival selection (ABSS) are proposed as two new operators for ANN pruning. The CBP operator retains significant neurons and prunes insignificant neurons on a probability basis and therefore prevents the exponential growth of an ANN. The ABSS operator can delete old ANNs with potentially complex structures and then introduce new ANNs with simple structures; thus, the ANNs are less likely to be trapped in a fully connected topology. The ECPA framework incorporates constructive and pruning approaches in an attempt to efficiently evolve compact ANNs. As a demonstration of the method, ECPA is applied to three prediction problems: the Mackey-Glass time series, the number of sunspots, and traffic flow. The numerical results show that ECPA makes the design of ANNs more feasible and practical for real-world applications. Crown Copyright (C) 2012 Published by Elsevier B.V. All rights reserved.
Applying Output Feedback Integral Sliding Mode Controller to Time-Delay Systems
(2011-04-01)
Ting, Huan-Chan
;
Chang, Jeang-Lin
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
For time-delay systems with mismatched disturbances and uncertainties, this paper developed an integral sliding mode control algorithm using output information only to stabilize the system. An integral sliding surface is comprised of output vectors and an auxiliary full-order compensator. The proposed output feedback sliding mode controller can satisfy the reaching and sliding condition and maintain the system on the sliding surface from the initial moment. When the specific linear matrix inequality has a solution, our method can guarantee the stability of the closed-loop system and satisfy the property of disturbance attenuation. Moreover, the design parameters of the controller and compensator can be simultaneously determined by the solution to the linear matrix inequality. Finally, a numerical example illustrated the applicability of the proposed scheme.
Applying Output Feedback Integral Sliding Mode Controller to Uncertain Time-Delay Systems with Mismatched Disturbances
(2011-12-01)
Ting, Huan-Chan
;
Chang, Jeang-Lin
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
For time-delay systems with mismatched disturbances and uncertainties, this paper develops an integral sliding mode control algorithm using output information only to stabilize the systems. An integral sliding surface is comprised of output signals and an auxiliary full-order compensator. The proposed output feedback sliding mode controller can satisfy the reaching and sliding condition and maintain the system on the sliding surface from the initial moment. When two specific algebraic Riccati inequalities have solutions, our method can guarantee the stability of the closed-loop system and satisfy the property of robust disturbance attenuation. Moreover, the design parameters of controller and compensator can be simultaneously determined by solutions to the algebraic Riccati inequalities. Finally, two numerical examples illustrate the applicability of the proposed scheme.
Design and performance analysis of an axial-flux disk-type switched reluctance motor for hybrid scooters
(2006-09-01)
Tsai, Jian-Feng
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
Switched reluctance motors (SRM) possess several superiorities such as simplicity in structure, ruggedness, and good reliability, which make it viable for the transportation applications. In this paper, an axial-flux disk-type switched reluctance motor (DSRM) for hybrid scooters is proposed. The DSRM can be used to drive the scooter and as an engine starter simultaneously. Moreover, the torque density of axial flux motors is more superior to that of radial flux motors in this application. Instead of utilizing FEM, the permeance model is employed to model the DSRM. With the analytical expressions, the output torque and current can be obtained dynamically and the geometries of the DSRM can be further optimized. Finally, the hardware-in-the-loop (HIL) simulation environment is constructed to perform full system simulation and the DSRM is shown to be capable of being an auxiliary power source for hybrid scooters.
Discrete time sliding-mode control design with grey
(2007-09-01)
Ting, Huan-Chan
;
Chang, Jeang-Lin
;
Yeh, Chih-Huei
;
Chen, Yon-Ping
;
電控工程研究所
This paper presents an advanced GM(1,1) model which can improve the accuracy of the conventional grey prediction and applies it to discrete sliding-mode control (DSMC). Using a Lagrange polynomial to take as a compensator and combining it with the original GM(1,1) model, the proposed prediction method can decrease the prediction error and easily implement in microprocessors with less computing time and memories. Then we employ this technique in DSMC to detect the system unknown perturbation. Comparing with the conventional DSMC, the proposed algorithm can reduce the switching gain and result in that the system state is bounded in a smaller region. Numeric simulation results of a DC motor are given to illustrate the feasibility and successfulness of the proposed design.
Face Liveness Detection Based on Skin Blood Flow Analysis
(2017-12-01)
Wang, Shun-Yi
;
Yang, Shih-Hung
;
Chen, Yon-Ping
;
Huang, Jyun-We
;
電子工程學系及電子研究所
Face recognition systems have been widely adopted for user authentication in security systems due to their simplicity and effectiveness. However, spoofing attacks, including printed photos, displayed photos, and replayed video attacks, are critical challenges to authentication, and these spoofing attacks allow malicious invaders to gain access to the system. This paper proposes two novel features for face liveness detection systems to protect against printed photo attacks and replayed attacks for biometric authentication systems. The first feature obtains the texture difference between red and green channels of face images inspired by the observation that skin blood flow in the face has properties that enable distinction between live and spoofing face images. The second feature estimates the color distribution in the local regions of face images, instead of whole images, because image quality might be more discriminative in small areas of face images. These two features are concatenated together, along with a multi-scale local binary pattern feature, and a support vector machine classifier is trained to discriminate between live and spoofing face images. The experimental results show that the performance of the proposed method for face spoof detection is promising when compared with that of previously published methods. Furthermore, the proposed system can be implemented in real time, which is valuable for mobile applications.
Face Recognition Based on Sparse Representation Applied to Mobile Device
(2014)
Chou, Kuan-Yu
;
Huang, Guan-Ming
;
Tseng, Hao-Chien
;
Chen, Yon-Ping
;
電控工程研究所
This paper develops an Android face recognition application for users on mobile device, and applies it in the face verification system of Samsung Galaxy SII smart phone. The developed face recognition application includes three parts, the face detection using the Viola-Jones face detection program, the feature extraction implemented by the eigenface features, and the face recognition based on the sparse representation of L2 norm minimization. Different to general learning methods, the developed application does not require a tremendous amount of time in data training and can achieve a high recognition rate even higher than 99%, for examples 99.2% for the Sheffield face database, 99.4% for the Cohn-Kanade face database and 96.5% for ORL face database. Finally, the face verification application proposed for the Samsung Galaxy SII smart phone indeed successfully verifies a face just in one second which makes it a real-time application.
Grey Neural Network-Based Forecasting System for Vision-Guided Robot Trajectory Tracking
(2011)
Yang, Shih-Hung
;
Chou, Chung-Hsien
;
Chung, Chen-Fang
;
Pai, Wen-Pang
;
Liu, Tse-Han
This paper presents a grey neural network-based forecasting system (GNNFS) in solving the prediction problem. GNNFS adopts a grey model to predict the signal and a neural network (NN) to forecast the prediction error of the grey model. A sequential batch learning (SBL) is developed to adjust the weights of the NN. The proposed GNNFS is applied to a binocular robot, called an Eye-Robot, for human-robot interaction which involved predicting the trajectory of a participant's hand and tracking the hand. By applying the SBL, the GNNFS can gradually learn to predict the trajectory of the hand and track it well. The experimental results show that the GNNFS can carry out the SBL in real-time for vision-guided robot trajectory tracking.
Integration of Grey Model and Neural Network for Robotic Application
(2011)
Yang, Shih-Hung
;
Li, Jung-Che
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This paper proposes an intelligent forecasting system based on a feedforward neural network aided grey model (FNAGM), integrating a first-order single variable grey model (GM(1,1)) and a feedforward neural network. The system includes three phases: initialization phase, GM(1,1) prediction phase, and FNAGM prediction phase. A number of parameters required for the FNAGM are selected in the initialization phase. A one-step ahead predictive value is generated in the GM(1,1) prediction phase, followed by the implementation of a feedforward neural network used to determine the prediction error of the GM(1,1) and compensate for it in the FNAGM prediction phase. We also adopted on-line batch training to adjust the network according to the Levenberg-Marquardt algorithm in real-time. According to the experimental results of a robot, the proposed intelligent forecasting system can provide high accuracy for both trajectory prediction and target tracking.
Intelligent Forecasting System Based on Grey Model and Neural Network
(2009)
Yang, Shih-Hung
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This paper presents the design issues of two intelligent forecasting systems, feedforward-neural-network-aided grey model (FNAGM) and Elman-network-aided grey model (ENAGM). Both he FNAGM and ENAGM combine a first-order single variable grey model (GM(1,1)) and a neural network (NN). The GM(1,1) is adopted to predict signal, and the feedforward NN and the Elman network in the FNAGM and ENAGM respectively are used to learn the prediction error of the GM(1,1). Simulation results demonstrate that the intelligent forecasting systems with on-line learning can improve the prediction of the GM(1,1) and can be implemented in real-time prediction.
Intelligent Forecasting System Using Grey Model Combined with Neural Network
(2011-03-01)
Yang, Shih-Hung
;
Chen, Yon-Ping
;
電機工程學系
;
電控工程研究所
;
Department of Electrical and Computer Engineering
This paper proposes an intelligent forecasting system based on a feedforward-neural-network-aided grey model (FNAGM), which integrates a first-order single variable grey model (GM(1,1)) and a feedforward neural network. There are three phases in the system process, including initialization phase, GM(1,1) prediction phase and FNAGM prediction phase. First, some parameters required in the FNAGM are chosen in the initialization phase. Then, a one-step-ahead predictive value is generated in the GM(1,1) prediction phase. Finally, a feedforward neural network is used to learn the prediction error of the GM(1,1) and compensate it in the FNAGM prediction phase. Significantly, an on-line batch training is adopted to adjust the network according to the Levenberg-Marquardt algorithm in real-time. From the simulation results, the proposed intelligent forecasting system indeed improves the prediction error of the GM(1,1) and obtains more accurate prediction than other numerical methods.
Maximum Power Point Tracking of Photovoltaic System Based on Reinforcement Learning
(2019-11-01)
Chou, Kuan-Yu
;
Yang, Shu-Ting
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
The maximum power point tracking (MPPT) technique is often used in photovoltaic (PV) systems to extract the maximum power in various environmental conditions. The perturbation and observation (P&O) method is one of the most well-known MPPT methods; however, it may face problems of large oscillations around maximum power point (MPP) or low-tracking efficiency. In this paper, two reinforcement learning-based maximum power point tracking (RL MPPT) methods are proposed by the use of the Q-learning algorithm. One constructs the Q-table and the other adopts the Q-network. These two proposed methods do not require the information of an actual PV module in advance and can track the MPP through offline training in two phases, the learning phase and the tracking phase. From the experimental results, both the reinforcement learning-based Q-table maximum power point tracking (RL-QT MPPT) and the reinforcement learning-based Q-network maximum power point tracking (RL-QN MPPT) methods have smaller ripples and faster tracking speeds when compared with the P&O method. In addition, for these two proposed methods, the RL-QT MPPT method performs with smaller oscillation and the RL-QN MPPT method achieves higher average power.
Multi-task Cascaded and Densely Connected Convolutional Networks Applied to Human Face Detection and Facial Expression Recognition System
(2019-01-01)
Chou, Kuan-Yu
;
Cheng, Yi-Wen
;
Chen, Wei-Ren
;
Chen, Yon-Ping
;
電控工程研究所
Face detection and recognition is an important issue and a difficult task in computer vision and human-computer interaction. Recently, with the development of deep learning, several related technologies have been proposed for face detection and facial expression recognition (FER), and the outstanding convolutional neural networks are the most common used in this field. This thesis applies the multi-task cascade convolutional neural network to face detection, and then designs the real-time FER system based on densely connected convolution network (DenseNet). The system first scales the input image to an image pyramid, and then uses the hierarchical network to determine whether a candidate window includes a human face. If a face exists, then send the candidate window to the FER system. Since DenseNet possesses the property of feature reuse, it can effectively reduce the amount of parameters and computation efforts, beneficial to develop the real-time system. In order to capture the variation of facial muscle in different expressions, this architecture adopts convolution operations with a stride 1 and tries different numbers of dense blocks. Through experiments, the proposed system can achieve real-time recognition in 30FPS and with recognition accuracy better than human eyes.
Neural Network Based Stereo Matching Algorithm Utilizing Vertical Disparity
(2010)
Yang, Shih-Hung
;
Ho, Cheng-Yu
;
Chen, Yon-Ping
;
電機工程學系
;
Department of Electrical and Computer Engineering
This paper presents a stereo matching algorithm utilizing vertical disparity (SMAVD) in solving the matching problem of stereo vision. SMAVD adopts a two-dimensional Hopfield neural network (HNN) to match the stereo pairs according to the energy function developed to describe three constraints including uniqueness, similarity and compatibility. The similarity of one matched pair is measured according to the difference of its neighboring pixels. The compatibility between two matched pairs is determined from not only smoothness and geometric comparisons but also vertical disparity comparison to improve the matching accuracy. Moreover, SMAVD uses a genetic algorithm to design the parameters of the nonlinear functions employed in the similarity and compatibility measures. By applying the updating rule, the HNN could obtain the correct matched pairs satisfying the constraints. The experimental results on the image pairs acquired from a binocular robot demonstrate that SMAVD could achieve high correct matching percentage with less computation time.
Output feedback integral sliding mode controller of time-delay systems with mismatch disturbances
(2012-01-01)
Ting, Huan-Chan
;
Chang, Jeang-Lin
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
For uncertain time-delay systems with mismatch disturbances, this paper presented an integral sliding mode control algorithm using output information only. An integral sliding surface is comprised of output signals and an auxiliary full-order compensator. The designed output feedback sliding mode controller can locally satisfy the reaching and sliding condition and maintain the system on the sliding surface from the initial moment. Since the system is in the sliding mode and two specific algebraic Riccati inequalities are established, the proposed algorithm can guarantee the stability of the closed-loop system and satisfy the property of disturbance attenuation. Moreover, the design parameters of the controller and compensator can be simultaneously determined by solutions to two algebraic Riccati inequalities. Finally, a numerical example illustrates the applicability of the proposed scheme.
Predictor design of a novel Grey model PGM21 using pseudo second-order information
(2006-06-01)
Li, Keh-Tsong
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
A novel grey model, called pseudo-GM(2, 1) or PGM21 for simplicity, is developed to improve the prediction accuracy for a non-monotone sequence by modifying the GM(1,1) model with a remedial term. Since the remedial term consists of two development coefficients, it can be treated as a kind of second-order information, called the pseudo second-order information in this paper. With the remedial term, the PGM21 only costs a little higher computation time than the original GM(1,1) model; however, its prediction accuracy is highly increased. The important features of the PGM21 are demonstrated by numeric simulation results.
Proportional-Derivative Unknown Input Observer Design Using Descriptor System Approach for Non-Minimum Phase Systems
(2011-10-01)
Ting, Huan-Chan
;
Chang, Jeang-Lin
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This paper considers the problem of estimating the state of an MIMO linear system with unknown inputs in the state and output. Through a series of linear transformations in the state and output equations, the original system can be transformed into a descriptor system form. The proposed proportional derivative observer can accurately estimate the system state and avoid the peaking phenomenon. Moreover, the approach developed in this paper does not require the derivatives of the output and can be applied to the system with unstable zeros (with respect to the relation between the output and the unknown input). Finally, our algorithm can prove the valid feasibility and the property of disturbance attenuation through demonstrating a simulation-base example.
Real-Time Fingerspelling Recognition System Design Based on RGB-D Image Information
(2014)
Chou, Kuan-Yu
;
Chuang, Hui-Chi
;
Chiu, Meng-Tzu
;
Chen, Yon-Ping
;
電控工程研究所
This paper provides a fingerspelling recognition system with high accuracy rate based on RGB-D image. The system are separated into three parts, including hand region detection, hand feature extraction, and fingerspelling recognition. For the hand regions detection, the regions of hand and face are first obtained by skin color detection and connect component labeling (CCL), and then the hand, the region-of-interest (ROI), is determined by the feature point extraction based on distance transform. Followed is the hand feature extraction which consists of the hand structure and the hand texture. From the feature points of ROI, the locations of palm and fingertips, palm direction, and finger vectors are formed as the hand structure. In addition to the hand structure, this paper adopts the LBP operator to generate the hand texture to deal with the fingerspelling not recognizable by the hand structure. Finally, the extracted hand features are sent into the fingerspelling recognition system, which is built with several different neural network classifiers. The experimental results show that this system is an effective real-time recognition system whose accuracy is higher than 80% for most of the fingerspelling in American Sign Language (ASL).
Symbiotic Neuron Evolution of a Neural-Network-Aided Grey Model for Time Series Prediction
(2011)
Yang, Shih-Hung
;
Chen, Yon-Ping
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This paper introduces a symbiotic neuron evolution algorithm (SNEA) to determine the topology of a neural-network-aided grey model (NNAGM) for time series prediction problem. The SNEA uses an evolutionary approach to evolve partially connected neural networks (NNs) and determine the number of hidden neurons. To achieve symbiotic evolution, SNEA first establishes a neuron population where each neuron is randomly created, and evaluates the neurons by constructing NNs with different numbers of neurons. Each neuron shares fitness from participating NNs. This algorithm then performs evolution on the neuron population by crossover and mutation based on neuron fitness. An NNAGM designed by SNEA is applied to the prediction problems and compared with other methods. The experimental results show that SNEA can produce an NNAGM with appropriate topology and higher prediction performance than other methods.
Symbiotic Structure Learning Algorithm for Feedforward Neural-Network-Aided Grey Model and Prediction Applications
(2017-01-01)
Yang, Shih-Hung
;
Huang, Wun-Jhu
;
Tsai, Jian-Feng
;
Chen, Yon-Ping
;
電子工程學系及電子研究所
The learning ability of neural networks (NNs) enables them to solve time series prediction problems. Off-line training can be applied to design the structure and weights of NNs when sufficient training data are available. However, this may be inadequate for applications that operate in real time, possess limited memory size, or require online adaptation. Furthermore, the structural design of NNs (i.e., the number of hidden neurons and connected topology) is crucial. This paper presents a novel algorithm, called the symbiotic structure learning algorithm (SSLA), to enhance a feedforward neural-network-aided grey model (FNAGM) for real-time prediction problems. Through symbiotic evolution, the SSLA evolves neurons that cooperate well with each other, and constructs NNs from the neurons with hyperbolic tangent and linear activation functions. During construction, the hidden neurons with the linear activation function can be simplified to a few direct connections from the inputs to the output neuron, leading to a compact network topology. The NNs share the fitness value with participating neurons, which are further evolved through neuron crossover and mutation. The proposed SSLA was evaluated through three real-time prediction problems. Experimental results showed that the SSLA-derived FNAGM possesses a partially connected NN with few hidden neurons and a compact topology. The evolved FNAGM outperforms other methods in prediction accuracy and continuously adapts the NN to the dynamic changes of the time series for real-time applications.