International Journal of Pattern Recognition and Artificial Intelligence

Distance metric is widely used in similarity estimation which plays a key role in fingerprint recognition. In this work we propose the detailed comparison of 29 distinct distance metrics. Features of fingerprint images are extracted using Fast Fourier Transform (FFT). Recognition rate, receiver operating curve (ROC), time and space complexity parameters are used for evaluation of each distance metric. To consolidate our conclusion we used the standard fingerprint database available at Bologna University and FVC2000 databases. After evaluation of 29 distinct distance metrics we found Sorgel distance metric performs best. Genuine acceptance rate (GAR) of Sorgel distance metric is observed to be ~5% higher than traditional Euclidean distance metric at low false acceptance rate (FAR). Sorgel distance gives good GAR at low FAR with moderate computational complexity.

Edge detection is often regarded as a basic step in range image processing by virtue of its crucial effect. The majority of existing edge detection methods cannot satisfy the requirement of efficiency in many industrial applications due to huge computational costs. In this paper, a novel instantaneous method, named RIDED-2D is proposed for denoising and edge detection for 2D scan line in range images. In the method, silhouettes of 2D scan line are classified into eight types by defining a few new coefficients. Several discriminant criteria on large noise filtering and edge detection are stipulated based on qualitative feature analysis on each type. Selecting some feature point candidates, a practical parameter learning method is provided to determine the threshold set, along with the implementation of an integrated algorithm by merging calculation steps. Because all the coefficients are established based on distances among the points or their ratio, RIDED-2D is inherently invariant to translation and rotation transformations. Furthermore, a forbidden region approach is proposed to eliminate interference of the mixed pixels. Key performances of RIDED-2D are evaluated in detail by including computational complexity, time expenditure, accuracy and stability. The results indicate that RIDED-2D can detect edge points accurately from several real range images, in which large noises and systematic noises are involved, and the total processing time is less than 0.1 millisecond on an ordinary PC platform using the integrated algorithm. Comparing with other state-of-the-art edge detection methods qualitatively, RIDED-2D exhibits a prominent advantage on computational efficiency. Thus, the proposed method qualifies for real-time processing in stringent industrial applications. Besides, another contribution of this paper is to introduce CPU clock counting technique to evaluate the performance of the proposed algorithm, and suggest a convenient and objective way to estimate the algorithm's time expenditure in other platforms.

This paper describes a fully automated computer vision system for detection and classification of the seven basic facial expressions using Multi-Class Support Vector Machine (SVM). Facial expressions are communicated by subtle changes in one or more discrete features such as tightening of the lips, raising the eyebrows, opening and closing of eyes or certain combination of them, which can be identified through monitoring the changes in muscle movements (Action Units), located around the regions of mouth, eyes and eyebrows. For classifying facial expressions, an analytic representation of face with 15 feature points has been used that provides visual observation of the discrete features responsible for the seven basic facial expressions. Feature points from the region of mouth are detected by segmenting the lip contour applying a variational formulation of the level set method. A multidetector approach of facial feature point detection is utilized for identifying the feature-points from the regions of eyes, eyebrows and nose. Feature vectors composed of 15 features are then obtained with respect to the average representation of neutral face and are used to train a Multiclass SVM classifier. The proposed method has been tested over two different facial expression image databases and the average successful recognition rates of 92.04% and 86.33% have been achieved.

An improved manifold learning method, called Uncorrelated Local Fisher Discriminant Analysis (ULFDA), for face recognition is proposed. Motivated by the fact that statistically uncorrelated features are desirable for dimension reduction, we propose a new difference-based optimization objective function to seek a feature submanifold such that the within-manifold scatter is minimized, and between-manifold scatter is maximized simultaneously in the embedding space. We impose an appropriate constraint to make the extracted features statistically uncorrelated. The uncorrelated discriminant method has an analytic global optimal solution, and it can be computed based on eigen decomposition. As a result, the proposed algorithm not only derives the optimal and lossless discriminative information, but also guarantees that all extracted features are statistically uncorrelated. Experiments on synthetic data and AT&T, extended YaleB and CMU PIE face databases are performed to test and evaluate the proposed algorithm. The results demonstrate the effectiveness of the proposed method.

In this paper, we propose a novel logic-rule learning approach for the Tower of Knowledge (ToK) architecture, based on Markov logic networks, for scene interpretation. This approach is in the spirit of the recently proposed Markov logic networks for machine learning. Its purpose is to learn the soft-constraint logic rules for labeling the components of a scene. In our approach, FOIL (First Order Inductive Learner) is applied to learn the logic rules for MLN and then gradient ascent search is utilized to compute weights attached to each rule for softening the rules. This approach also benefits from the architecture of ToK, in reasoning whether a component in a scene has the right characteristics in order to fulfil the functions a label implies, from the logic point of view. One significant advantage of the proposed approach, rather than the previous versions of ToK, is its automatic logic learning capability such that the manual insertion of logic rules is not necessary. Experiments of labeling the identified components in buildings, for building scene interpretation, illustrate the promise of this approach.

LDA (Latent Dirichlet Allocation) proposed by Blei is a generative probabilistic model of a corpus, where documents are represented as random mixtures over latent topics, and each topic is characterized by a distribution over words, but not the attributes of word positions of every document in the corpus. In this paper, a Word Position-Related LDA Model is proposed taking into account the attributes of word positions of every document in the corpus, where each word is characterized by a distribution over word positions. At the same time, the precision of the topic-word's interpretability is improved by integrating the distribution of the word-position and the appropriate word degree, taking into account the different word degree in the different word positions. Finally, a new method, a size-aware word intrusion method is proposed to improve the ability of the topic-word's interpretability. Experimental results on the NIPS corpus show that the Word Position-Related LDA Model can improve the precision of the topic-word's interpretability. And the average improvement of the precision in the topic-word's interpretability is about 9.67%. Also, the size-aware word intrusion method can interpret the topic-word's semantic information more comprehensively and more effectively through comparing the different experimental data.

Temperature modulation has been proved to be an efficient technique for improving the selectivity and stability of gas sensors. In this paper, a new signal processing approach is proposed for metal oxide gas sensor signals under the modulation of its operating temperature, which combined a novel global feature extraction method based on the Hilbert–Huang Transform with a pattern recognition method based on neural network. By using the empirical mode decomposition method, the dynamic signals are decomposed into the intrinsic modes that coexist in the sensor system, and a better understanding of the nature of the gas sensing response information contained in the sensor response signals is approached. The method is demonstrated by an application in the identification and quantification of gas mixtures containing three flammable species using a micro gas sensor. The three gas analytes are methane (CH₄), ethanol (C₂H₆O) and carbon monoxide (CO). And the relative average quantification errors for the three gases are about 7%, 8% and 12%, respectively.

Besides security, trust is another key factor affecting resource sharing. The progress and existing problems of trust management related works were analyzed, and several key factors for impacting trust construction were pointed out. Trust management from the aspect of attribute-based recommended trust, trust improvement, trust domain and trust evaluation is investigated. Under certain conditions, a member can improve its direct and recommended trust by initiative cooperation and thus obtain more resource sharing opportunities. These methods were then used in new-member joining problems. With the debit and credit method, a small group with related benefit was established, and at the same time new-member joining, behavior monitoring, resource sharing competence were encouraged.