Catálogo de publicaciones - libros

This paper describes, using situational awareness and Genetic algorithm, a run-time optimization methodology of the Gabor wavelet parameters so that it produces a feature space for efficient object recognition. Gabor wavelet efficiently extracts the feature space of orientation selectivity, spatial frequency and spatial localization. Most previous object recognition approaches using Gabor wavelet do not include systematic optimization of the parameters for the Gabor kernel, even though the system performance might be much sensitive to the characteristics of the Gabor parameters. This paper explores efficient object recognition using adaptive Gabor wavelet based situational aware method. The superiority of the proposed system is shown using IT-Lab, FERET and Yale face database. We achieved encouraging experimental results.

This paper proposes an evolvable hardware system with capability of evolution under uneven image environment, which is implemented on reconfigurable field programmable gate array (FPGA) platform with ARM core and genetic algorithm processor (GAP). Parallel genetic algorithm based reconfigurable architecture system evolves image filter blocks to explore optimal configuration of filter combination, associated parameters, and structure of feature space adaptively to uneven illumination and noisy environments for an adaptive image processing. The proposed evolvable hardware system for image processing consists of the reconfigurable hardware module and the evolvable software module, which are implemented using SoC platform board with the Xilinx Virtex2 FPGA, the ARM core and the GAP. The experiment result shows that images affected by various environment changes are enhanced for various illumination and salt & pepper noise image environments.

We propose a method for an efficient frontal face detection using skin color, integrated feature space, and post processing. The proposed method reduces the search space by facial color information and detects face candidate windows by integrated feature space. The integrated feature space consists of intensity and texture information. Multiple Bayesian classifiers are employed for selection of face candidate windows on integrated feature space. And we use face and face-like nonface samples to training these Bayesian classifiers. Finally, face regions of the detected candidates are selected by merging and filtering post processing.

This paper proposes an Active Queue Management(AQM) scheme referred to as Fuzzy Predictive Preferential Dropping (FPPD). Two contributions are made in this work. Firstly, a fuzzy predictor is employed to improve the accuracy of traffic prediction. Secondly, a novel congestion index, predicted traffic intensity from fast flows, is used to derive packet dropping probability for AQM. The FPPD safely detects real congestion caused by large flows while leaving transient traffic burst from short-lived flows alone. Furthermore, a preferential dropping mechanism is adopted to treat packets from long-term fast flows and short-lived flows differently. Simulations show that the proposed FPPD reduces packet drop ratio and utilizes link bandwidth more efficiently than other AQM schemes. It also improves the quality of service perceived by web users.

DEA (data envelopment analysis) is a non-parametric technique for measuring and evaluating the relative efficiencies of a set of decision-making units (DMUs) in terms of a set of common inputs and outputs. Traditionally, the data of inputs and outputs are assumed to be measured with precision, i.e., the coefficients of DEA models are crisp value. However, this may not be always true. There are many circumstances where precise inputs and outputs can not be obtained. Under such situations, data of inputs and outputs can be represented by fuzzy numbers. Based on the dual program of DEA models, we propose fuzzy DEA models for CCR and BCC models. Our fuzzy DEA models provide crisp efficiency with fuzzy input and output data.

From a viewpoint of integrating control and scheduling, the impact of resource availability constraints on the implementation of iterative optimal control (IOC) algorithms is considered. As a novel application in the emerging field of feedback scheduling, fuzzy technology is employed to construct a feedback scheduler intended for anytime IOC applications. Thanks to the anytime nature of the IOC algorithm, it is possible to abort the optimization routine before it reaches the optimum. The maximum iteration number within the IOC algorithm is dynamically adjusted to achieve a desired CPU utilization level. Thus a tradeoff is done between the available CPU time and the quality of control. Preliminary simulation results argue that the proposed approach is effective in managing the inherent uncertainty in control task execution and delivers better performance than traditional IOC algorithm in computing resource constrained environments.

Significant efforts in developing active queue management (AQM) in gateway routers in a TCP/IP network have been made since random early detection (RED) in 1993, and most of them are statistical based. Our approach is to capitalize on the understanding of the TCP dynamics to design an effective AQM scheme. In this paper, two FL-based AQM algorithms are proposed with the deployment of traffic load factor for early congestion notification. Extensive experimental simulations with a range of traffic load conditions have been done for the purpose of performance evaluation. The results show that the proposed two FLAQM algorithms outperform some well-known AQM schemes in terms of both user-centric measures and network-centric measures.

This paper proposes an iris pattern recognition algorithm as one of biometric techniques applied to identify a person using his/her physiological characteristics. Since the iris pattern of human eye has an unique and invariant texture, we can use it as a biometric key. First, we obtain the feature vector from the fuzzy LDA after performing 2D Gabor wavelet transform. And then, we compute the similarity measure based on the correlation. Here, since we use four matching values obtained from four different directional Gabor wavelets and select the maximum value among them, it is possible to reduce the recognition error. To show the usefulness of the proposed algorithm, we applied it to an iris database consisting of 300 iris patterns extracted from 50 subjects and finally got more higher than 90% recognition rate.

In our previous work [7], we presented a robust centroid target tracker based on new distance features in cluttered image sequences. A real-time adaptive segmentation method based on new distance features was proposed for the binary centroid tracker. The target classifier by the Bayes decision rule for minimizing the probability error should properly estimate the state-conditional densities. In this correspondence, the proposed target classifier adopts the fuzzy-reasoning segmentation using the fuzzy membership functions instead of the estimation of the state-conditional probability densities. Comparative experiments also show that the performance of the proposed fuzzy- reasoning segmentation is superior to that of the conventional thresholding methods. The usefulness of the method for practical applications is demonstrated by considering two sequences of real target images. The tracking results are good and stable without difficulty of the estimation.

In this paper, the aim is to solve the effect of a filter on an input signal by using fuzzy logic. We used the characteristic of lowpass filter for this application. If the input frequencies are very high, it will be reduced. We obtained a rule-base for emulating this characteristic. Two methods are used for this problem. In the first method, fuzzy inputs are signal values and the differences of the signal values. This is called “the difference method”. The other approach is to calculate the average of two discrete values of the signal. As known, averaging is simple lowpass filter which smoothes out high frequency variations in a signal. But fuzzy approach is used for averaging of the signal. And this is called “the average method”.