統語的パターンマッチングによる作業動作認識

概要

論文の詳細を見る
Work motion recognition is an important function for computer support systems to provide novice workers autonomous instruction by monitoring human behavior. Since manual work, such as assembling work, is performed by seamless motion at variable speed, it needs special solution in pattern matching problems. This paper proposes one type of syntactic pattern matching method for motion recognition in human work. The proposed algorithm has two steps of pattern matching. The first process is to divide a time series of motion data into a relatively short time motion by defined conditions, and to assign a symbol of motion type to the potion data. The second one is to compare the symbol stream with work models In this process, Dynamic Programming (DP) matching method is used. Experimental results show that the proposed method is effective to adapt the variable work speed even if short time motion model is used.
2003-12-05