複雜運動系統的學習曲線

The learning curves of complex motor skill system

根據動力系統理論，複雜運動可視為一個多元的動力系統。分析系統的時間刻度可以了解其行為之演進；指數函數固定的時間刻度代表系統接近穩定狀態，而複雜系統中連續或重疊的時間刻度可能出現對數函數。目的：本研究的目的，在透過分析學習曲線的時間刻度來探討複雜運動技能的學習過程，以了解複雜運動學習曲線的意義。方法：研究中紀錄四名實驗參加者學習兩手丟三球技能過程中的接球數和運動學資料，以指數函數和對數函數就接球數曲線和時間結構曲線進行適配，並且對動作協調子系統進行主成分分析。結果：整體的接球數曲線呈現對數函數行為，而分段結果曲線中發現的指數函數與技能表現具有一致性。時間結構曲線經逐次改變率的檢查雖不支持指數函數，但適配值也與技能表現達到吻合。經過學習，代表動作協調的成分數由3個增加到4至5個，主要成分的解釋量也有顯著的變化（F（2，21）＝3.99*）。結論：複雜運動技能的學習過程涵蓋多元子系統的演進，因此在整體性的學習曲線上呈現對數函數，而分析子系統的學習函數則可以進ㄧ步了解子系統的互動關係。Complex motor skill is considered a multiform dynamics according to dynamical system theory. To gain insight into the evolution of this integration, one can analyses the changing rate of the systems. The constant time scale of exponential function characterizes the behavior of approaching the stable state while the power function implicates the continuous and aggregate time scales. Purpose: the purpose of this study was to investigate the process of learning the complex motor skill by examining the time scale of learning curves. Method: the numbers of catch and the kinematics data of four participants learning the three-ball cascading juggling were recorded for analyses. The learning curves of catching and temporal structure were fitted using exponential function and power function. For recognizing the pattern of movement coordination, the principal component analysis was applied. Results: the power function behavior was shown on the ball-catching curves of whole practice sessions. However, the behavior showed in the segment curves of ball-catching corresponded to skill performance. Although the curves of temporal structure were explained reasonably by the exponential function and with the consistent behavior with the skill performance, this result did not supported by Rn test. The number of component which stands for the movement coordination increased from three to four or five over practice. The percentage of explanation changed significantly (F(2, 21)=3.99*) as well. Conclusion: the process of complex motor skill integrated the evolution of time scales of multiple subsystems and demonstrated the behavior of power function for overall leaning curves. Further knowledge of the interaction among subsystems could be found by investigating learning function.