Rifat Edizkan

Subspace based feature selection for pattern recognition

Feature selection is an essential topic in the field of pattern recognition. The feature selection strategy has a direct influence on the accuracy and processing time of pattern recognition applications. Features can be evaluated with either univariate approaches, which examine features individually, or multivariate approaches, which consider possible feature correlations and examine features as a group. Although univariate approaches do not take the correlation among features into consideration, they can provide the individual discriminatory power of the features, and they are also much faster than multivariate approaches. Since it is crucial to know which features are more or less informative in certain pattern recognition applications, univariate approaches are more useful in these cases. This paper therefore proposes subspace based separability measures to determine the individual discriminatory power of the features. These measures are then employed to sort and select features in a multi-class manner. The feature selection performances of the proposed measures are evaluated and compared with the univariate forms of classic separability measures (Divergence, Bhattacharyya, Transformed Divergence, and Jeffries-Matusita) on several datasets. The experimental results clearly indicate that the new measures yield comparable or even better performance than the classic ones in terms of classification accuracy and dimension reduction rate.

The common vector approach and its comparison with other subspace methods in case of sufficient data

This paper presents an application of the common vector approach (CVA), an approach mainly used for speech recognition problems when the number of data items exceeds the dimension of the feature vectors. The calculation of a unique common vector for each class involves the use of principal component analysis. CVA and other subspace methods are compared both theoretically and experimentally. TI-digit database is used in the experimental study to show the practical use of CVA for the isolated word recognition problems. It can be concluded that CVA results are higher in terms of recognition rates when compared with those of other subspace methods in training and test sets. It is also seen that the consideration of only within-class scatter in CVA gives better performance than considering both within- and between-class scatters in Fishers linear discriminant analysis. The recognition rates obtained for CVA are also better than those obtained with the HMM method.

Development of indoor positioning system with ultrasonic and infrared signals

Position measurement is one of the primal problems in wide range of applications such as air/land/marine vehicles tracking, automation systems, and equipment tracking in medical sector, robotics and sensor networks. Location information in outdoor application can be easily obtained from well-known systems such as GPS and GLONASS. Since GPS signal is not available in indoor area, several positioning systems have been developed for. But there is not a consensus for indoor positioning systems like the outdoor. In this study, a positioning system that uses ultrasonic and infrared signals is designed and implemented. The system is tested in 4m2 coverage range and it is observed that the maximum absolute error is less than +/-2 cm.

Use of Novel Feature Extraction Technique with Subspace Classifiers for Speech Recognition

Speech recognition is one of the fast moving research areas in pervasive services requiring human interaction. Like any type of pattern recognition system, selection of the feature extraction method and the classifier play a crucial role for speech recognition in terms of accuracy and speed. In this paper, an efficient wavelet based feature extraction method for speech data is presented. The feature vectors are then fed into three widely used linear subspace classifiers for recognition analysis. These classifiers are Class Featuring Information Compression (CLAFIC), Multiple Similarity Method (MSM) and Common Vector Approach (CVA). TI-DIGIT database is used to evaluate the performance of speaker independent isolated word recognition system designed. Experimental results indicate that the proposed feature extraction method together with the CLAFIC and CVA classifiers give considerably high recognition rates.

Estimation of adaptive neuro-fuzzy inference system parameters with the expectation maximization algorithm and extended Kalman smoother

In this paper, we propose a new supervised learning method for adaptive neuro-fuzzy inference system (ANFIS) training, which uses the expectation maximization (EM) algorithm and extended Kalman smoother (EKS) together; we refer to it here as the EM-EKS training method. While the EKS tunes the ANFIS parameters, the EM algorithm estimates the parameters of the Kalman filter and avoids non-optimal performance. Besides, we also propose a new algorithm to select the initial values of the EKS parameters. We compare the EM-EKS method of ANFIS training with traditional ANFIS training. Although the new training method requires more computing time, it yields improved RMSE values in function approximation and prediction problems. Examples of benchmark function approximation and prediction illustrate the effectiveness of the EM-EKS ANFIS training method.

A survey of indoor localization systems

Nowadays, indoor and outdoor location-aware applications has been become increasingly widespread. Location information for outdoor applications is obtained from well-known systems such as GPS and GLONASS. Although there are several systems for indoor localization, there is not a consensus on them like the outdoor. In this study, several aspects of indoor localization systems such as system topologies, mesaurement techniques, location calculation methods, performance metrics, signal types are overviewed. Some foresights are also made about the future of indoor localization.

Automatic face recognition from frontal images

Face recognition can be described as identification of people from their face images. In this study, an automatic face recognition system has been designed by using frontal images photographed in our lab. The automatic face recognition procedure consists of an alignment process which includes face detection, eye detection, mapping of the center coordinates of the eyes to a standard face template. This is followed by classification of aligned faces. In literature, face alignment process is usually done with manually and high recognition rates can be achieved due to very well aligned faces. However, in real-time face recognition applications, its not possible to align face images manually. Therefore, successful classification rates reported in the literature are mostly misleading. In this study, we aligned faces in a fully automatic manner and we obtained more reliable and realistic face recognition rates. Face images are represented with gray level, LBP, LTP, and two dimensional Gabor filter features and performances are tested with Eigenfaces, Fisherfaces, and DCV methods. Experimental results showed that the automatic recognition rates can reach close to 90% correct recognition rates.

Comparisons of features for automatic eye and mouth localization

Localization of the eyes and mouth in face images is very important for accurate classification in automatic face recognition systems. The alignment of unknown face images with templates generally improves the performance of the face recognition system, and this process uses locations of the eyes and mouth. In this work, we compare different features (gray-level values, distance transform features, gradients and local binary patterns) for automatic localization of eyes and mouth. To this end, we use the sliding window approach using the linear and nonlinear support vector machine (SVM) classifiers. We created new frontal face data sets to train and test our algorithms. The experimental results show that the SVM classifier using the Gaussian kernel yields better results than the linear kernel. Among the four feature extraction methods, the performance of the local binary pattern features draws the attention for having better detection rates in both the linear and the nonlinear cases with smaller feature size.

Performance improvement in distributed Turkish continuous speech recognition system using packet loss concealment techniques

In this work, we investigated performance improvement of the distributed Turkish continuous speech recognition system (TCSRS) with some well-known packet loss concealment (PLC) techniques. The PLC techniques, Lagrange, Spline and maximum a-posteriori (MAP) are applied to the sparse and burst packet losses in the system. The experimental results showed that the interpolation methods give acceptable performance during sparse packet losses. But for burst losses, the performance of MAP estimation method is better than that of interpolation methods.

Evaluation of image descriptors in subspace-based classifiers for traffic sign recognition

In this study, the performance of some image descriptors in traffic sign recognition is obtained using the subspace-based classifiers. The subspace methods make both dimension reduction in feature space and maximize the classification rate. The feature vectors are extracted from the images containing a traffic sign by image descriptors. Gray scale, Histogram of Oriented Gradients (HOG), Local Binary Patterns (LBP), and Local Phase Quantization (LPQ) are used as image descriptors in our study. The feature vectors are processed by the subspace methods, Principle Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Discriminative Common Vector (DCV), for recognizing traffic signs. In the experimental study, the database containing triangular and circular signs was used. The database also includes shifted and rotated traffic signs. The recognition performances of the subspace-based classifiers were compared with the template matching method. The best classification performances are obtained for the HOG features and DCV method. The classification rates for triangular and circular signs are 98.38% and 99.25% respectively.