Improvement of Generalization Ability of Kernel-Based Fisher Discriminant Analysis for Recognition of Japanese Sign Language Hand Pastures, "Yubi-Moji", Using K-means Method

Clustering of the training data using K-means method improved the generalization ability of kernel-based Fisher discriminant analysis. In this study, Gaussian type kernel function was used. The proposed method is applied to the vision-based recognition of the 41 Japanese Sign Language (JSL) static hand postures, which express part of the Japanese syllabary. The reflective near-infrared light method was used with few burdens for a hand postures hand postures in which the outline were similar cannot be discriminated, or a hand had to be colored. In this study, the reflective near-infrared light method without a cumbersome glove-like device is applied to the recognition of the 41 JSL static hand postures. After preprocessing for the raw data, feature vector was extracted by K-means method and Gaussian type kernel function [5]. For classification of 41 postures of 4 subjects, linear discriminant analysis (LDA) was used. expressioner compared with the conventional methods. The image data obtained by this 2 Manual Japanese Syllabary,

[1]  Vladimir Pavlovic,et al.  Visual Interpretation of Hand Gestures for Human-Computer Interaction: A Review , 1997, IEEE Trans. Pattern Anal. Mach. Intell..

[2]  Takio Kurita,et al.  Scale Invariant Face Detection and Classification Method Using Shift Invariant Features Extracted from Log-Polar Image , 2001 .

[3]  Neural Networks for Signal Processing VIII , 1998, Neural Networks for Signal Processing VIII. Proceedings of the 1998 IEEE Signal Processing Society Workshop (Cat. No.98TH8378).

[4]  B. Scholkopf,et al.  Fisher discriminant analysis with kernels , 1999, Neural Networks for Signal Processing IX: Proceedings of the 1999 IEEE Signal Processing Society Workshop (Cat. No.98TH8468).

[5]  Kenji Nishida,et al.  A Topographic Kernel-based Regression Method , 2002, JCIS.