摘要：Human body action recognition technology is a challenging problem in Human-computer interaction. The computer vision and motion measurements are two primary recognition methods. And the action recognition includes dynamic action recognition and gesture recognition. Acceleration sensor-based human action recognition technology belongs to the dynamic action recognition. This paper proposed an interval overlapping degree extension method in upper limbs motion recognition based on acceleration median (IODE-AM). Stretch arm has the feature that the arm’s acceleration increases firstly, and then decreases. Therefore, the proposed method chooses the acceleration median of the arm outstretching process and the direction of acceleration at the initial moment of arm outstretching as its recognition characteristic. It can reduce the affection of outstretching speed to the characteristic value of limb action, and can achieve the goal that the different outstretching speed actions having same direction could be described by the same characters. Then combining the extension pattern recognition method based on the interval overlapping degree, the IODE-AM recognized the limb action. The experiment results show that the recognition accuracy rate of the IODE-AM is 93.2 %.
