Self-Organization for Temporal Data of Varying Length
Gesture recognition is becoming popular, because it is an interesting domain with both temporal and spatial character. There have been two approaches in self-organization for temporal or sequence data. One is to regard data at each time as an input to self-organization. The other is to regard temporal or sequence data as an input to self-organization. In this paper we propose to use multiple Gaussian functions to efficiently compress information of temporal data in the latter approach. It is robust to noise and temporal variation, and also holds enough information. This idea is simple, but is effective in dealing with temporal data in self-organization. Experiments using artificial and real temporal data have demonstrated the effectiveness of the proposed method compared to linear interpolation in terms of recognition rate and the mean quantization error.
Masumi Ishikawa Hiroshi Suenaga
Dept. of Braia Science and Engineering Graduate School of Life Science and Systems Engineering Kyushu Institute of Technology, 680-4 Kawazu, lizuka, Fukuoka 820-8502, Japan
国际会议
8th International Conference on Neural Information Processing(ICONIP 2001)(第八届国际神经信息处理大会)
上海
英文
285-290
2001-11-14（万方平台首次上网日期，不代表论文的发表时间）