Ayşegül Uçar

A new facial expression recognition based on curvelet transform and online sequential extreme learning machine initialized with spherical clustering

In this paper, a novel algorithm is proposed for facial expression recognition by integrating curvelet transform and online sequential extreme learning machine (OSELM) with radial basis function (RBF) hidden node having optimal network architecture. In the proposed algorithm, the curvelet transform is firstly applied to each region of the face image divided into local regions instead of whole face image to reduce the curvelet coefficients too huge to classify. Feature set is then generated by calculating the entropy, the standard deviation and the mean of curvelet coefficients of each region. Finally, spherical clustering (SC) method is employed to the feature set to automatically determine the optimal hidden node number and RBF hidden node parameters of OSELM by aim of increasing classification accuracy and reducing the required time to select the hidden node number. So, the learning machine is called as OSELM-SC. It is constructed two groups of experiments: The aim of the first one is to evaluate the classification performance of OSELM-SC on the benchmark datasets, i.e., image segment, satellite image and DNA. The second one is to test the performance of the proposed facial expression recognition algorithm on the Japanese Female Facial Expression database and the Cohn-Kanade database. The obtained experimental results are compared against the state-of-the-art methods. The results demonstrate that the proposed algorithm can produce effective facial expression features and exhibit good recognition accuracy and robustness.

The solutions of vibration control problems using artificial neural networks

Abstract This paper introduces an alternative method artificial neural networks (ANN) used to obtain numerical solutions of mathematical models of dynamic systems, represented by ordinary differential equations (ODEs) and partial differential equations (PDEs). The proposed trial solution of differential equations (DEs) consists of two parts: The initial and boundary conditions (BCs) should be satisfied by the first part. However, the second part is not affected from initial and BCs, but it only tries to satisfy DE. This part involves a feedforward ANN containing adjustable parameters (weight and bias). The proposed solution satisfying boundary and initial condition uses a feedforward ANN with one hidden layer varying the neuron number in the hidden layer according to complexity of the considered problem. The ANN having appropriate architecture has been trained with backpropagation algorithm using an adaptive learning rate to satisfy DE. Moreover, we have, first, developed the general formula for the numerical solutions of n th-order initial-value problems by using ANN. For numerical applications, the ODEs that are the mathematical models of linear and non-linear mass-damper-spring systems and the second- and fourth-order PDEs that are the mathematical models of the control of longitudinal vibrations of rods and lateral vibrations of beams have been considered. Finally, the responses of the controlled and non-controlled systems have been obtained. The obtained results have been graphically presented and some conclusion remarks are given.

Modelling and simulation with neural and fuzzy-neural networks of switched circuits

Recently, the modelling and simulation of switched systems containing new nonlinear components in electronics and power electronics industry have gained importance. In this paper, both feed‐forward artificial neural networks (ANN) and adaptive network‐based fuzzy inference systems (ANFIS) have been applied to switched circuits and systems. Then their performances have been compared in this contribution by developed simulation programs. It has been shown that ANFIS require less training time and offer better performance than those of ANN. In addition, ANFIS using “clustering algorithm” to generate the rules and the numbers of membership functions gives a smaller number of parameters, better performance and less training time than those of ANFIS using “grid partition” to generate the rules. The work not only demonstrates the advantage of the ANFIS architecture using clustering algorithm but also highlights the advantages of the architecture for hardware realizations.

Object recognition and detection with deep learning for autonomous driving applications

Autonomous driving requires reliable and accurate detection and recognition of surrounding objects in real drivable environments. Although different object detection algorithms have been proposed, not all are robust enough to detect and recognize occluded or truncated objects. In this paper, we propose a novel hybrid Local Multiple system (LM-CNN-SVM) based on Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs) due to their powerful feature extraction capability and robust classification property, respectively. In the proposed system, we divide first the whole image into local regions and employ multiple CNNs to learn local object features. Secondly, we select discriminative features by using Principal Component Analysis. We then import into multiple SVMs applying both empirical and structural risk minimization instead of using a direct CNN to increase the generalization ability of the classifier system. Finally, we fuse SVM outputs. In addition, we use the pre-trained AlexNet and a new CNN architecture. We carry out object recognition and pedestrian detection experiments on the Caltech-101 and Caltech Pedestrian datasets. Comparisons to the best state-of-the-art methods show that the proposed system achieved better results.

Moving towards in object recognition with deep learning for autonomous driving applications

Object recognition and pedestrian detection are of crucial importance to autonomous driving applications. Deep learning based methods have exhibited very large improvements in accuracy and fast decision in real time applications thanks to CUDA support. In this paper, we propose two Convolutions Neural Networks (CNNs) architectures with different layers. We extract the features obtained from the proposed CNN, CNN in AlexNet architecture, and Bag of visual Words (BOW) approach by using SURF, HOG and k-means. We use linear SVM classifiers for training the features. In the experiments, we carried out object recognition and pedestrian detection tasks using the benchmark the Caltech 101 and the Caltech Pedestrian Detection datasets.

A new formulation for classification by ellipsoids

We propose a new formulation for the optimal separation problems. This robust formulation is based on finding the minimum volume ellipsoid covering the points belong to the class. Idea is to separate by ellipsoids in the input space without mapping data to a high dimensional feature space unlike Support Vector Machines. Thus the distance order in the input space is preserved. Hopfield Neural Network is described for solving the optimization problem. The benchmark Iris data is given to evaluate the formulation.

Illumination Invariant Face Recognition Using Principal Component Analysis – An Overview

Illumination variation is a challenge problem at face recognition since a face image varies as illumination changes. In this paper, it is reviewed the illumination variation methods in the state-of-the-art such as the single scale retinex algorithm, the multi scale retinex algorithm, the gradientfaces based normalization method, the Tan and Triggs normalization method and the single scale weberfaces normalization method. The face recognition is performed by using Principal Component Analysis (PCA) in MATLAB environment. AR face database is used for evaluating the face recognition algorithm using PCA. The distance classifier called as Squared Euclidean is used. Experimental results are comparatively demonstrated.

Fuzzy model identification using support vector clustering method

We have observed that the support vector clustering method proposed by Asa Ben Hur, David Horn, Hava T. Siegelmann, Vladimir Vapnik, (Journal of Machine Learning Research, (2001), 125-137) can provide cluster boundaries of arbitrary shape based on a Gaussian kernel abstaining from explicit calculations in the high-dimensional feature space. This allows us to apply the method to the training set for building a fuzzy model. In this paper, we suggested a novel method for fuzzy model identification. The premise parameters of rules of the model are identified by the support vector clustering method while the consequent ones are tuned by the least squares method. Our model does not employ any additional method for parameter optimization after the initial model parameters are generated. It gives also promising performances in terms of a large number of rules. We compared the effectiveness and efficiency of our model to the fuzzy neural networks generated by various input space-partition techniques and some other networks.

Person Recognition via Facial Expression Using ELM Classifier Based CNN Feature Maps

Extreme learning machine (ELM) and deep learning methods are well-known with their efficiency, accuracy, and speed. In this study, we focus on the application of ELM to a deep learning structure for person recognition with facial expressions. For this purpose, a new convolutional neural network (CNN) model containing Kernel ELM classifiers was constructed. In this model, ELM was not used only as a fully connected layer replacement and energy function was employed to generate feature maps for the ELM. There are two advantages of the proposed model. First, it is fast and successful in face recognition studies. Second, it can drastically improve the performance of a partially-trained CNN model. Consequently, the proposed model is very suitable for CNN models, where the learning process requires a lot of time and computational power. The model is tested with the Grimace data set and experimental results are presented in details.

Project based learning using student clubs and short online videos, in-class activities

This paper presents a project based learning approach applied to the teaching of the Fundamentals of Electrical and Electronics Engineering (FEEE). The training is composed of several modules such as the in-class activities, practical student activities or courses organized by student clubs, learning materials, term project usable as product in real life, short exams, and homework assignments. The proposed approach allows students to better understand circuit components and circuit principles. In this paper, three projects are presented as examples of project based learning using student clubs. The students successfully completed their all projects. Especially they generated a specific prototype for the term projects. All results were evaluated and success of the students were presented.