Recognition of Behavioral Patterns
Performance in sports depends largely on the athlete’s coordination with his co-players, opponents, and objects
Recognition of Behavioral Patterns
Resumen
Performance in sports depends largely on the athlete’s coordination with his co-players, opponents, and objects (e.g., a ball) as well as his/her abilities to extract the relevant information to do so. We propose a functional architecture motivated by concepts of dynamical systems theory for the control as well as recognition of behavioral patterns. That is, action and perception are conceived of as two side of the same coin. We view behavior as patterns evolving in phase flows as these provide an unambiguous presentation of the temporal evolution of dynamical processes. In addition, phase flow topologies allow for the unambiguous classification of dynamical systems—and consequently of their modeled behavior: the behavior of two systems map onto each other if and only if their phase flow is topologically equivalent. For our present purpose, we illustrate our approach focusing on 2-dimensional systems, although the concept holds for systems of arbitrary dimension. In 2-dimensional systems, the phase space is spanned by position and velocity, in which case the only possible topological structures are point attractors and limit cycles that are associated with discrete behavior (and postures) and rhythmic behaviors, respectively (so-called separatrices may also exist). Whenever the architecture is exposed to an evolving trajectory, it is matched against phase flows it has in ‘memory’ such that over time the flow with the closest correspondence wins (i.e., is recognized), which we numerically exemplify. Behavioral patterns may be recognized before a movement has finished, in which case the architecture may be said to anticipate.
