DYNAMIC PATTERN-RECOGNITION OF COORDINATED BIOLOGICAL MOTION
DYNAMIC PATTERN-RECOGNITION OF COORDINATED BIOLOGICAL MOTION
复制标题
DOI：
10.1016/0893-6080(90)90022-d
复制
发表时间：
1990-01-01
期刊：
影响因子：
7.8
通讯作者：
PANDYA, AS
中科院分区：
文献类型：
--
作者：
HAKEN, H;KELSO, JAS;PANDYA, AS
我们开发了一种算法，能够识别和分类的视觉创建的模式协调的生物运动，特别是不同的多关节肢体轨迹产生的人类。该算法使用已识别的集体变量或序参数作为编码这些模式的基础，这些模式是通过Kelso及其同事对相变的实验研究获得的。因此，有意义的信息识别动态视觉模式驻留在吸引子的序参数动力学。在神经网络中，这些序参数代表了网络整体的不同宏观状态（而不是单个神经元的相互作用），从而以类似完形过程的方式构成了一个协同组织的神经场。
We develop an algorithm that enables the identification and categorization of visually created patterns of coordinated biological motion, specifically, different multijoint limb trajectories produced by humans. The algorithm uses identified collective variables or order parameters as the basis for encoding these patterns, obtained through experimental studies of phase transitions by Kelso and colleagues. Thus, meaningful information for recognizing dynamic visual patterns resides in attractors of the order parameter dynamics. In a neural net, these order parameters represent different macrostates of the net as a whole (rather than the interactions of single neurons), thereby constituting a synergetically organized neural field, in the fashion of a Gestalt-like process.