To improve the human-computer interaction, it is very important to estimate the user's intention by analyzing the human behaviors such as the gesture and the gazing. The goal of our study is to detect the face orientation as the primary information at least several times per second and less than five-degree estimation error. In this paper,we propose the real-time algorithm to recognize the face orientation. The basic concept of the algorithm is to combine the simple and compensational methods. Three types of algorithm are shown as examples that compensate each other and analyzed the compensational characteristics. Based on the analysis, the combination formula is deduced. It includes the compensation of the central position and the direction data and the weighted-averaging of the plural output data. The experiment shows that the algorithm can successfully recognize the vertical and the horizontal orientation of the face at two times a second on the personal computer and the effect of the compensation is large.

View full abstract

The sociometry analysis applying fuzzy node fuzzy graph shows the social structure. But, a fuzzy node fuzzy graph is usually complicated, so, we proposed a method to transform a fuzzy node fuzzy graph to a crisp node fuzzy graph by using T-norm. We shall explain a fuzzy node fuzzy graph, its analysis method and its effective application to a fuzzy sociometry analysis by using new T-norm "Quesi-Logical product".

View full abstract

The nearest neighbor (NN) methods solve classification problem by storing examples as points in a feature space, which requires some means of measuring distances between examples. However, it suffers from the existence of noisy attributes. One resolution is to modify the distance of similarity degree using attribute weights, which can not only decrease the influence of noisy attributes, but also subset relevant attributes. In this paper,a rough genetic algorithm (RGA) proposed by Lingras and Davies is applied to the classification problem under an undetermined environment, based on a fuzzy distance function by calculating attribute weights. The RGA can complement the existing tools developed in rough computing. Computational experiments are conducted on benchmark problems, downloaded from UCI machine learning databases. Experimental results,compared with a usual GA[1] and the C4.5 algorithms, verify the efficiency of the developed algorithm. Furthermore, the weights learned by the proposed learning method is applicable to not only fuzzy similarity functions but also any similarity functions. As an application, a new distance metric, weighted discretized value difference metric (WDVDM), is proposed. Experimental results show that the WDVDM improves the discretized value difference metric (DVDM).

View full abstract

An instance-based algorithm to restore the color information of both still and dynamic images with high gradation is proposed. The 175 pairs of the color information (L^*a^*b^* colorimetric system) in color scheme cards under low and standard illumination are used as the instances. The color information of a pixel in the input image is restored with the most relevant instance to the color information of the pixel, by adding the difference between the low and the standard color information of the instance. The relevancy of an instance to the input pixel is determined by the color-difference based score, which is modified by considering its neighborhood pixels within a frame and the pixels calculated by optical flow between the previous frame and the processed frame. Experimental results using the surveillance system with CCD camera show that the proposed algorithm attains better performance in terms of the noise removal and color difference (about 10% shorter) compared with the conventional instancebased algorithm. The proposed algorithm gives the foundation for a surveillance camera system.

View full abstract

Fuzzy modeling based on fuzzy ID3 has been used for its efficiency in selecting input variables. One of major difficulties of fuzzy ID3 is that it is difficult to determine the fuzzy partition in case human's sufficient knowledge about the partition is not available. Against this difficulty, a method for fuzzy partition based on information contained in data was proposed. This method, utilizes the shape of expected value of information, and does not need veteran's experience. However the degree of freedom in deciding the membership functions for the fuzzy partition is high, and the method shows only a guideline for the decision. This paper presents a method for determining the membership functions for fuzzy ID3 based on smoothing of acquired information. The membership functions are to be used for a fuzzy model. This method utilizes the unique feature of fuzzy technology that is smoothing of data and interpolation of inferred values of rules. The proposed method is applied to identification of predictor of oxygen component of an oil refinery plant.

View full abstract