Mitsu Yoshimura

Analysis and evaluation of dancing movement based on LMA

The final goal of this research is to extract characteristic poses as well as highlight parts from data of dancing movement obtained by motion capturing technique. For this, the theory of Laban movement analysis (LMA) has been applied, and the physical feature values corresponding to the LMA components are defined. By observing the change over time of these feature values, body movements corresponding to the LMA components are extracted. In this paper we mainly focus on effort and shape components of LMA. Results have been compared with those which have been extracted by a LMA specialist.

Off-line verification of Japanese signatures after elimination of background patterns

A system of off-line signature verification composed of a preprocessing stage and a verification stage is proposed in this paper. It is assumed that each signature is written on a surface with a background pattern and a sample of the background pattern is available. The preprocessing stage to eliminate the background pattern from a signature image consists of 5 steps: the position adjustment between a signature image and a background image, the filtering of the background pattern by local maximization, the clipping of random noises from the background pattern, the background elimination by transformation of the multilevel signature image to a binary signature image through subtraction and thresholding, and the smoothing for defects recovery and isolated noise elimination. The effect of the preprocessing operation is illustrated in figures. The verification stage following the preprocessing stage is based on the Arc Pattern Method proposed in Ref. 7. An experiment is performed to examine the performance of the proposed system. Twenty Japanese autographs, for which 20 authentic signatures, 10 genuine counter signatures and 10 forged counter signatures are provided, are used in the experiment. The experiment achieved an error rate of as low as approximately 14%.

Quantification and recognition of basic motion okuri in Japanese traditional dance

This work investigates a method for quantifying and recognizing two basic motions, feminine and description, of okuri in Japanese traditional dance. Index variables for quantifying typical form of dance motions are defined and measured on dancing played by a distinguished expert, with results indicating the effectiveness for discrimination of motion types. The indices are effective also for discriminating the movements of two dancers who played dancing with a common scenario different from the model dancing by the expert. Although the effectiveness is dependent on various factors such as the quality of motion samples, the subspaces and distance measures used in the multivariate analysis, the individuality of dancers and uniqueness of dance motions are well recognized using these indices in an experiment.

Writer identification based on the arc pattern transformation

A method of writer identification for handwritten characters that is based on a series of arc-pattern transformations, defined as transformations from character patterns to frequency patterns of arcs with given chords and curvatures, is proposed. The method is an extension of that due to K. Shakunaga et al. (1984). Its effectiveness is examined experimentally using copies of a Japanese sentence with 25 letters, written by hand by each of 24 persons. About 2% average increase in correct identification rate is achieved by this method compared with the method of Shakunaga et al.<<ETX>>

Authorship Identification of Ukiyoe by Using Rakkan Image

This paper describes a method of identifying authorship of Ukiyoe prints by using Rakkan images found in the prints. A weighted direction index histogram method has been used to create the feature vector for Rakkan character analysis. Also the Pseudo Mahalanobis distances were used to judge distances between dictionary templates and test data. The method includes binarization of Rakkan images which has been done by recursively applying the Otsu method, which realized good character segmentation performance, eliminating the influence of stain and smears. We used 100 Ukiyoe prints which were drawn by 10 artists as both templates and test samples. The identification experiment was done using the leave-one-out method. The results of the experiment indicate there is a possibility of successfully using this method for identifying authorship of Ukiyoe.

Comparison of Structural Variables with Spatio-temporal Variables Concerning the Identifiability of Okuri Class and Player in Japanese Traditional Dancing

This paper experimentally examined the characteristics of structural variables and spatio-temporal variables through the identifiability of Okuri classes and players (dancers) in traditional Japanese dancing. The data for the experiment were acquired using a motion capture system, which monitored dancing motions performed by a master player (dancer) followed by two expert dancers and three beginners. The discrimination of two Okuri classes, feminine Okuri and descriptive Okuri, and that of 5 dancers was experimentally attempted using structural variables and spatio-temporal variables. The experimental result suggested that spatio-temporal variables were more advantageous than structural variables for identification of Okuri classes and dancers

Statistical displacement analysis for handwriting verification

In this paper, it is assumed that each writer has his or her own statistics of handwriting displacement, therefore a statistical displacement analysis for handwriting verification is proposed. Here, a regularization method with the coarse-to-fine strategy computes the displacement function in questionable handwritten letters, and then it is normalized to remove the noisy displacement that arises from the position drift and scaling variation. Finally, the normalized displacement function and the statistics of displacement obtained in advance from registered authentic letters are used to calculate the distance from a standard handwritten letter to a questionable one. A fundamental simulation was conducted in order to evaluate the performance of the proposed method.

Evaluation of Signature Quality as a Function of Nationality Via an Off-Line Signature Verification System

Abstract This paper investigates differences in signature quality as a function of nationality by comparing Japanese and Indonesian signatures. A signature verification system based on the localized arc pattern method was applied to one Japanese and two Indonesian (educated and low-educated) signature data sets collected by the authors. As the observed error rate for the Japanese data set was as high as 7.2% compared to only 4.0% for the educated Indonesian set, this indicates Japanese signatures are less robust against forgery. We hypothesize that Japanese cultura/social habits are responsible for this, as Japanese people do not accustomed to use their signature as a means of identification.