International Journal of Pattern Recognition and Artificial Intelligence

Many locality-based unsupervised dimensionality reduction (DR) algorithms have recently been proposed and demonstrated to be effective to a certain degree in some classification tasks. In this paper, we aim to show that: (1) a discriminant disposal is intentionally or unintentionally induced from the construction of locality in these unsupervised algorithms, however, such a discrimination is often inconsistent with the actual class information, so here called disguised discrimination; (2) sensitivities of these algorithms to local neighbor parameters stem from the inconsistency between the disguised discrimination and the actual class information; (3) how such inconsistency impacts the classification performance of these algorithms. The experiments on the benchmark face datasets testify our statements that are expected to provide some insight into the unsupervised leaning based on locality.

A novel approach for detecting video logo-removal forgery is proposed by measuring inconsistency of blur. Our approach is based on the assumption that if a digital video undergoes logo-removal forgery; the blurriness of the forged region is expected to be different as compared to the nontampered parts of the video. Blurriness is first estimated by analyzing the spatial and temporal statistical property of logo areas, and suspicious areas are roughly located; then features are extracted and a fine classification is implemented by applying support vector machine (SVM) to extract features. If the suspicious areas and the reference areas are classified into different classes, the video is judged as a forged video. Experimental results show that our method is robust to video lossy compression for logo-removal forgery detection with the advantages of high classification accuracy and low computation cost.

This paper, presents a novel unsupervised dimensionality reduction approach called variance difference embedding (VDE) for facial feature extraction. The proposed VDE method is derived from maximizing the difference between global variance and local variance, so it can draw the close samples closer and simultaneously making the mutually distant samples even more distant from each other. VDE utilizes the maximum variance difference criterion rather than the generalized Rayleigh quotient as a class separability measure, thereby avoiding the singularity problem when addressing the sample size problem. The results of the experiments conducted on ORL database, Yale database and a subset of PIE database indicate the effectiveness of the proposed VDE method on facial feature extraction and classification.

This study has developed an object detection and segmentation technique for processing cytoplasm and cell nucleus on ThinPrep-cervical smear images at various magnifications. Both edge detection techniques and region growing for adaptive threshold were applied to a segment cell nucleus, a cytoplasm, and backgrounds using a cervical cell image.

To validate the accuracy and feasibility of the proposed method, we took a variety of cervical cell images to perform a series of experiments. The images were of superficial cells, intermediate cells, and abnormal cells, with each taken from ThinPrep smears at various magnifications. The results indicate that the proposed method can automatically segment cell nucleus and cytoplasm regions while accurately extracting object contours. These results can serve as a reference for examiners of cell pathologies.

Data mining currently is a vital and core process because of advances in the technology, and numerous other forms of data. As an essential and significant constituent of data mining, classification has also become vital and a sought-after-option for research. Over the years, as various soft computing tools have been proven to be more complimentary than competitive, many hybrid approaches have evolved for pattern classification. A major thrust has always been there to apply hybrid approaches involving artificial neural network as the main tool for pattern classification. Here, one such hybrid approach is presented with application of Rough set philosophy to unsupervised artificial neural network (UANN) based pattern classifier. This enhances the performance of the classifier and makes it more suitable for real time classification applications dealing with large and unlabeled data. The idea presented here discusses with possibility and application of various Rough set based approaches at different levels of implementation of such classifier. The concluding section presents the results achieved for data of two case studies related to face recognition and mobile handoff prediction.

In this paper, we propose a uniform enhancement approach called smoothing function method, which can cooperate any optimization algorithm and improve its performance. The method has two phases. In the first phase, a smoothing function is constructed by using a properly truncated Fourier series. It can preserve the overall shape of the original objective function but eliminate many of its local optimal points, thus it can well approach the objective function. Then, the optimal solution of the smoothing function is searched by an optimization algorithm (e.g. traditional algorithm or evolutionary algorithm) so that the search becomes much easier. In the second phase, we switch to optimize the original function for some iterations by using the best solution(s) obtained in phase 1 as an initial point (population). Thereafter, the smoothing function is updated in order to approximate the original function more accurately.

These two phases are repeated until the best solutions obtained in several successively second phases cannot be improved obviously. In this manner, any optimization algorithm will become much easier in searching optimal solution. Finally, we use the proposed approach to enhance two typical optimization algorithms: Powell direct algorithm and a simple genetic algorithm. The simulation results on ten challenging benchmarks indicate the proposed approach can effectively improve the performance of these two algorithms.

Modeling and analyzing the dynamic shape of human motion is a challenging task owing to temporal variations in the shape and multiple sources of observed shape variations such as viewpoint, motion speed, clothing, etc. We present a new framework for dynamic shape analysis based on temporal normalization and factorized shape style analysis. Using a nonlinear generative model with motion manifold embedding in a low-dimensional space, we detect cycles of periodic motion like gait in different views and synthesize temporally-aligned shape sequences from the same type of motion at different speeds. The bilinear analysis of temporally-aligned shape sequences decomposes dynamic motion into time-invariant shape style factors and time-dependent motion factors. We extend the bilinear model into a tensor shape model, a multilinear decomposition of dynamic shape sequences for view-invariant shape style representations. The shape style is a view-invariant, time-invariant, and speed-invariant shape signature and is used as a feature vector for human identification. The shape style can be adapted to new environmental conditions by iterative estimation of style and content factors to reflect new observation conditions. We present the experimental results of gait recognition using the CMU Mobo gait database and the USF gait challenging database.

In the present work we address the problem of phonetic segmentation of emotional speech. Investigating various traditional and recent HMM-based methods for speech segmentation, which we elaborated for the specifics of emotional speech segmentation, we demonstrate that the HMM-based method with hybrid embedded-isolated training offers advantageous segmentation accuracy, when compared to other HMM-based models used so far. The increased precision of the segmentation is a consequence of the iterative training process employed in the hybrid-training method, which refines the model parameters and the estimated phonetic boundaries taking advantage of the estimations made at previous iterations. Furthermore, we demonstrate the benefits of using purposely-built models for each target category of emotional speech, when compared to the case of one common model built solely from neutral speech. This advantage, in terms of segmentation accuracy, justifies the effort for creating and employing the purposely-built segmentation models per emotion category, since it significantly improves the overall segmentation accuracy.